Abstract
Although it is well established that TLR9 recognizes CpG-DNA, the structural details of ligand-receptor interaction are still mostly unknown. The extracellular domain of TLR9 is composed of 25 leucine-rich repeat (LRR) motifs, 5 of which bear inserting sequences that do not conform to the LRR consensus motif. In this study, we show that the functional integrity of the extracellular domain of murine TLR9 is lost by deletion of individual LRR motifs. When deleting only the inserting sequences, we observed that LRR2, 5, and 8 contribute to receptor activation by CpG-DNA. The latter deletions did not affect receptor dimerization but inhibited CpG-DNA binding. On the basis of a homology modeling approach, we furthermore identify a positively charged region in the N terminus that is essential for CpG-DNA-induced TLR9 activation. This interaction site mirrors findings previously shown for the structural recognition of dsRNA by TLR3 and hints toward a general principle of nucleic acid recognition by the respective TLR.
Innate immunity relies on TLR to detect invading microorganisms (1). TLR are germline-encoded type I integral membrane glycoproteins whose extracellular domain (ECD)4 is responsible for ligand binding. The ECD of TLR is composed of 19–25 leucine-rich repeats (LRR) (2). Each LRR forms a loop in which conserved hydrophobic residues point inward, and several of these loops build-up a horseshoe-shaped ECD, which is N- and C-terminally flanked by so-called cysteine flanking regions, termed LRR-NT and -CT, respectively. The first 10 aa of LRR are conserved in all LRR subtypes and form a β-sheet shaping the concave surface of the ECD, whereas the remaining portion of the LRR is variable among the different subtypes and forms the convex surface. All TLR contain varying numbers of “irregular” LRR that do not entirely conform to the consensus motif but bear inserting stretches of amino acids, which protrude from the horseshoe-shaped backbone and have been proposed to be involved in ligand binding (3).
To date, three modes of ligand binding by mammalian TLR have been identified (4). In the TLR1/TLR2 heterodimer the acyl chains of the Pam3CSK4 ligand are directly inserted into hydrophobic channels stretching LRR9–12 (5). For TLR4, LPS is presented to the receptor through a binding protein, MD-2 (6). In contrast, the crystal structure of TLR3 bound to dsRNA showed two binding sites for RNA that are formed by charged patches on its surface (7, 8, 9). One is located in the N-terminal part of the ECD (involving the LRR-NT and LRR1–3), and the other one is located in the C terminus with the irregular LRR20 contributing to RNA binding. Mutational studies confirmed the binding of RNA to residues in a region encompassing LRR20 and deletion of the insertion in LRR20 led to a complete loss of function (10).
In contrast to other TLR, the nucleic acid recognizing TLR3, 7, 8, and 9 recognize their ligands in intracellular compartments such as endosomes. A further noticeable feature of TLR7–9 is that the irregular LRR of these receptors are located at the same positions and that the insertions are very homologous (3). Another similarity between these receptors is the presence of a less structured region between LRR14 and 15 with low similarity to the LRR consensus. It has been proposed that this region may bring flexibility to the receptor (11). In contrast to TLR3, which recognizes dsRNA sequence independently (12), TLR7–9 are stimulated by nucleic acids in a sequence-specific manner (13, 14), which implies that in addition to the general binding characteristics of TLR3 for nucleic acids, recognition by TLR7–9 may be more complex. Surprisingly, for TLR9, it has been recently shown that cleavage of the ECD occurs and it was proposed that the C-terminal fragment starting from LRR15 mediates ligand recognition (15, 16).
Whereas receptor dimerization and subsequent signal transduction occur upon ligand binding for TLR2-TLR1 and TLR3, it was published for TLR9 that this receptor exists as a preformed dimer (11). Binding of stimulatory DNA was proposed to result in a conformational change in the receptor, which decreases the diameter of the ECD. This process was suggested to bring the TIR domains in close proximity, thereby activating downstream signaling. In this study, we tested the hypothesis that insertion-bearing LRRs within TLR9 contribute to ligand recognition. Additionally, we sought to determine whether specific binding sites in the TLR9 ECD can be mapped.
Materials and Methods
Cells and reagents
CpG-oligodeoxynucleotide (ODN) no. 1668 (TCCATGACGTTCCTGATGCT) was custom synthesized by MWG-Biotech either 3′-biotinylated or unmodified. Resiquimod (R848) was purchased from InvivoGen. LPS from Salmonella minnesota was provided by U. Seydel (Research Center Borstel, Borstel, Germany). DMEM, RPMI 1640 medium, and FCS were purchased from Biochrom. Abs were received from Cell Signaling Technology. Immobilized protein A was obtained from Thermo Scientific, streptavidin-agarose was from Prozyme, and the cathepsin inhibitor z-FA-fmk was from Biovision. RAW 264.7 cells, a murine macrophage cell line, were a gift from R. Schumann (Institute for Microbiology and Hygiene, Berlin, Germany). HEK293 cells were obtained from S. Bauer (Institute for Immunology, Marburg, Germany).
Site-directed mutagenesis
mTLR9-HA-plasmid was constructed by replacing the YFP-tag in TLR9-YFP (obtained by T. Espevik, Institute of Cancer Research and Molecular Medicine, Trondheim, Norway) with an HA-tag, using the restriction enzymes XhoI and Bsp1407I and custom synthesized 5′-phosphorylated ODN encoding the hemagglutinin (HA)-tag sequence YPYDVPDYA. The same approach was used to obtain a TLR9-myc-plasmid (encoded sequence: EEQKLISEEDL). Site-directed mutagenesis was conducted using primers containing the desired mutations using the QuikChange kit from Stratagene. Sequences of the primers can be made available upon request. Several clones were subjected to automated sequencing to confirm the mutation and to exclude further undesired mutations.
Reporter gene experiments
For reporter gene experiments a firefly luciferase reporter construct with a 6× NF-κB responsive element was used. HEK293 cells (0.2 × 106) were transfected in 24-well format and a volume of 500 μl. mTLR9-HA (25 ng) or the indicated mutant plasmid or 100 ng of mTLR7 plasmid (obtained from S. Bauer) together with 25 ng of NF-κB-reporter plasmid encoding Firefly luciferase and 100 ng pRL-TK (Promega) encoding Renilla luciferase were transfected using Lipofectamine 2000 (Invitrogen). Forty-eight hours after transfection, cells were stimulated, and luciferase activities were determined an additional 6 h later using the Dual Luciferase Reporter Assay System Kit (Promega). Shown are mean values of duplicates of one of at least three independent experiments.
Western blotting
HEK293 cells were transfected in a 6-well format with 4 μg of the indicated plasmid. Forty-eight hours later, cells were lysed for 30 min on ice in 250 μl of lysis buffer (50 mM Tris-HCl (pH 7.4), 1% Igepal, 0.25% sodium deoxycholate, 150 mM NaCl, 1 mM EDTA, 1 mM PMSF, 1 μg/ml each of aprotinin, leupeptin, and pepstatin, 1 mM Na3VO4, and 1 mM NaF). Lysates were cleared by centrifugation at 4°C for 15 min at 11,000 × g. Equal amounts of lysates were fractionated by 8% SDS-PAGE and electrotransferred to polyvinylidene difluoride membranes. The membranes were blocked with TBS (pH 7.8)/5% nonfat dry milk/0.05% Tween 20 and were blotted with the indicated abs according to the manufacturer’s protocol. HRP-labeled Abs were visualized by ECL (Amersham Biosciences).
Pull-down assay
HEK293 cells were transfected with TLR9-HA wild type (WT) or the indicated mutant plasmid and lysed as described above. Lysates were incubated with 5 μM 3′-biotin-CpG-ODN at 4°C for 1.5 h. Subsequently, 20 μl of streptavidin-agarose beads was added for 60 min. Beads were washed five times in lysis buffer, and eluted proteins were used for Western blot analysis (17). HRP-labeled Abs were visualized by ECL (Amersham Biosciences) using the Chemi-Smart 2000 video documentation system from PEQLAB Biotechnologie in the automated mode to avoid overexposure and saturation of the signal. Quantification of the detected signals was done using the Bio-1D Advanced-Software (PEQLAB Biotechnologie). The amount of precipitated TLR9 was normalized to the amount of receptor in the lysate and compared with the TLR9 WT pull-down.
Immunoprecipitation
A total of 1 × 106 HEK293 cells was cotransfected with 4 μg of TLR9-myc and 4 μg of TLR9-HA WT or the indicated mutant in a volume of 1 ml medium. A total of 200 μl of the lysates of these cells was incubated with 3 μl of anti-myc Ab for 3 h at 4°C. After further incubation with 20 μl of protein A beads for 2 h, precipitates were washed five times, and eluted proteins were used for Western blot analysis.
Determination of TNF-α and NO secretion
A total of 1.5 × 105 RAW264.7 cells was preincubated with the indicated amount of cathepsin inhibitor z-FA-FMK for 2 h and then stimulated with various TLR ligands. After 24 h of stimulation, TNF-α was determined in cell-free supernatants by ELISA, according to the manufacturer’s instruction, and fresh media were added to the cells. Other 24 h later supernatants were analyzed for NO accumulation photometrically (550 nm) by mixing equal parts of supernatant and Griess reagent (1:1 mixture of 1% sulfanilamide/5% H3PO4 and 0.1% naphthyl-ethylenediamine dihydrochloride).
Homology modeling and structure analysis
Homology modeling of mouse TLR9 ECD was conducted as previously described (18) based on the published human TLR3 ECD crystal structures (7, 19). In brief, the TLR9 ECD was generated stepwise by modeling N- and C-terminal subdomains and LRR14 individually using the MODELLER package (20). These were manually assembled, and spatial violations resulting from the manual docking procedure were corrected using GROMACS molecular dynamics (21). The complete ECD model was then scored for energy and sterical correctness using the ANOLEA (22), VERIFY_3D and ERRAT (http://nihserver.mbi.ucla.edu/) online servers. Structure analysis was conducted in SwissPBD Viewer (23) and PyMol (www.pymol.org) visualization software using PDB2PQR (24), PropKa (25), and APBS (26) packages for charge surface calculation.
Results
The aim of this study was to identify regions within the ECD of TLR9 involved in receptor activation. We first tested the hypothesis that irregular LRRs bearing inserting sequences that do not conform to the consensus LRR motif might play a role in receptor activation. In murine TLR9 such sequences can be found in LRR2, 5, 8, and 11 (insertion at position 10, numbering according to Ref. 3) and in LRR20 (after position 15) as shown in Fig. 1⇓A.
Deletion or exchange of complete LRR in TLR9 abolishes receptor function. A, Homology model of the TLR9 ectodomain with irregular LRR2, 5, 8, 11, 20 highlighted in red and regular LRR6, 16 in blue. B and C, HEK293 cells were transfected with the indicated TLR9 mutant plasmids, stimulated with 1 μM CpG-ODN 48 h later, and analyzed for NF-κB reporter gene activity (mean ± SD, one of four experiments). D, Dimer formation was assessed by immunoprecipitation with myc Ab and subsequent analysis with HA-tag-specific Abs by Western blot analysis (∗, possible cleavage product of TLR9).
Deletion of single LRR modules abolishes TLR9 activity
Initially, we made single deletions of each irregular LRR and tested the mutants (termed Del_LRRx, x denotes number of LRR; Table I⇓) in an NF-κB-dependent reporter assay in HEK293 cells. Deletion of any complete LRR bearing the inserting sequences was not tolerated and led to loss of function of the receptor (Fig. 1⇑B). Only in the case of LRR11, some residual activity remained. Surprisingly, even the deletion of LRR16, an entirely regular LRR, which was included as a control, resulted in loss of receptor activity. We concluded that this approach influences the overall structure of the horseshoe shape of the ECD, thus interfering with receptor activity irrespective of the position of the LRR.
Generated constructs of TLR9
On the basis of the high homology between TLR9 and TLR7 with regard to the position and composition of the irregular LRR, we next switched LRR2, 5, 8, 11, and 20 from TLR9 to those from TLR7 (sLRRx; Table I⇑). Again, all of the mutants featuring only one switched LRR module showed loss of responsiveness toward the TLR9 stimulus CpG-ODN (Fig. 1⇑C). Corroborating the above findings, even the switch of an LRR without insertion (sLRR16) or switching a highly homologous LRR (sLRR6; see also below) led to loss of responsiveness. This suggests an intricate spatial arrangement of LRR in the TLR9 ECD whose disruption results in the loss of receptor function. Not surprisingly, chimeras where all LRR containing irregular insertions of TLR9 were replaced with the corresponding ones of TLR7 (LRR2, 5, 8, 11, and 20) also showed loss of CpG-ODN responsiveness (data not shown). Moreover, the latter mutant did not become responsive to stimulation with the TLR7 ligand R848 (data not shown). Thus, it was not possible to switch ligand specificity by exchanging the irregular LRR modules between TLR7 and TLR9, which contrasts findings for TLR1 and TLR6, where switching of different LRR modules led to an exchange in ligand specificity of these two receptors (27). The correct expression of all above mutant constructs was confirmed by Western blotting and flow cytometry of YFP-tagged mutants (Fig. 1⇑D and data not shown). All tested mutants were expressed at comparable levels and at the expected size (130 kDa).
Deletion of single LRR modules does not affect TLR9 dimer formation
In contrast to TLR3, TLR9 is present as preformed dimers (11). To analyze whether the above mutants could still form such dimers, we conducted coimmunoprecipitations of two differently tagged TLR9s. No stimulation was performed, as preformed dimers were assumed to exist. Deletion of any complete LRR in the region between LRR-NT and LRR20 had no impact on the formation of preformed dimers (Fig. 1⇑D). Coimmunoprecipitations were specific as TLR4-HA was not detected in precipitates from cells cotransfected with both TLR4-HA and TLR9-myc. The results were in concordance with data from TLR3 where only regions in the very C-terminal part of the ECD participated in direct protein-protein interactions (LRR22, 23, and LRR-CT) (9). When detecting TLR9 by Western blotting, we noted that in addition to the full-length protein a second protein of ∼95 kDa was detected. In accordance with two recent reports on cleavage of TLR9 (15, 16), this band would correspond to an N-terminally truncated TLR9. Of note, none of the mutations appeared to lead to a loss of the cleavage product.
N-terminally located irregular LRR contribute to CpG-ODN binding and TLR9 activation
To examine only the influence of the insertions in the irregular LRR and in order not to disturb the structure by changing the number of LRR modules in the ECD, we generated mutants which lacked only the amino acids that did not conform to the consensus LRR-motif. Removal of insertions in the N-terminal LRR2, 5, and 8 (Del_Insx; Table I⇑) led to a complete loss in receptor function (Fig. 2⇓A). In contrast, deleting the insertions in LRR11 and LRR20 did not (Del_Ins11) or only partially (Del_Ins20) inhibit TLR9 activation, indicating that these insertions are not involved in ligand recognition and receptor activation and that in general deletion of individual insertions may be tolerated. Correct expression of the mutant proteins was verified by Western blotting, FACS analysis, and confocal microscopy (Fig. 2⇓D and data not shown).
Removal or modification of LRR insertions interferes with TLR9 activity. HEK293 cells were transfected with the indicated TLR9 plasmids either alone (A and C) or together with untagged TLR9 WT (1:50 ratio) (B) or myc-TLR9 WT (D). Forty-eight-hour posttransfection cells were stimulated, and NF-κB-dependent luciferase activity was measured 6 h poststimulation (A and B). C, Binding of biotinylated CpG-ODN to HA-tagged TLR9 was determined by streptavidin-agarose pull-down and quantitative Western blotting. D, Dimer formation was assessed by immunoprecipitation with myc Ab and subsequent analysis with HA-tag-specific Abs by Western blot analysis (∗, possible cleavage product of TLR9; mean ± SD of at least two (D) or three (A–C) independent experiments).
Additionally, we analyzed whether this set of mutants also exerted a dominant-negative effect when coexpressed together with TLR9 WT and stimulated with CpG-ODN (Fig. 2⇑B). Deletion of the insertions in LRR2, 5, and 8 but not in LRR11 and 20 exhibited a dominant-negative effect when coexpressed with TLR9 WT. Whereas Del_Ins20 showed no changes in NF-κB-activation cotransfection of Del_Ins11 with TLR9 WT even increased the activation, comparable to cotransfection of empty vector with TLR9 WT. Thus, Del_Ins11 behaved similar to TLR9 WT.
We next sought to address whether the mutants were still able to bind to CpG-ODN. Therefore, HEK293 cells were transfected with the respective mutants or TLR9 WT and the receptor was precipitated from protein lysates using biotinylated CpG-ODN (17). The amount of precipitated protein was normalized to the amount of protein present in the lysate and compared with precipitated TLR9 WT (Fig. 2⇑C). Of note, this kind of binding assay showed a considerable level of unspecific binding (here to TLR4-HA) probably due to the nature of the phosphothioate modification in CpG-ODN (28). Therefore, precise image quantification was conducted to enable a comparative analysis. For mutants lacking the specific insertions in the irregular LRR (Del_Insx), only Del_Ins11 and 20 retained the ability to bind CpG-ODN, in line with the activity measured earlier (Fig. 2⇑A). Conversely, the insertions in LRR2, 5, and 8 were involved in ligand binding (Fig. 2⇑C). Interestingly, all mutants lacking insertions were able to engage in dimer formation with TLR9 WT (Fig. 2⇑D) as observed before for the mutants with deletions of whole LRR. These findings indicate that the insertions are not involved in receptor-receptor interaction but are important for ligand recognition.
Surface charge is important for TLR9 activation by CpG-DNA
Having established that individual LRR insertions contribute to TLR9 function, we sought to address if distinct patches in the N-terminal part of the TLR9 ECD contribute to receptor function. A three-dimensional homology model of the mTLR9 ectodomain was generated (18). LRR6 is the most conserved LRR between TLR9 and 7 (Fig. 3⇓A). The first 10 residues of this LRR (part 6_1; Fig. 3⇓A) are the ones that have the same consensus pattern in all LRR subforms and form the β-sheets building the concave surface of the horseshoe-shaped ECD. Replacement of these first 10 residues was tolerated (Fig. 3⇓C). Unexpectedly, the remaining 11 aa (part 6_2; Fig. 3A⇓) could not be exchanged, even though only 4 of the 11 residues differ between the two receptors (position amino acid TLR9/TLR7: 211L/V, 213K/A, 216R/T, and 217Q/T). However, surface charge was drastically altered when replacing residues in part 2 of TLR9 with those from TLR7 (Fig. 3⇓B). Mutation of only one of the four differing residues did not affect receptor function (Fig. 3⇓C). Only for L211V less absolute activity was measured, but due to a lower control activation, this equaled to an identical x-fold induction. Changing individual residues in our model did not greatly affect surface charge (data not shown).
LRR6 is required for signal transduction due to a positively charged surface patch. A, Stick representation of LRR6 with part 1 (residues 200–209) colored in orange and part 2 (residues 210–221) colored green. The sequence alignment of murine TLR9 and 7 is shown for LRR6. B, Surface charge computation of LRR6 part 2 before change of TLR9 residues 210–221 (WT LRR6; top panel) and after change to TLR7 residues 214–225 (sLRR6; bottom panel) with residues of interest in TLR9 labeled. Red, negative charge; blue, positive charge. C, HEK293 cells were transfected with the indicated TLR9 plasmids, stimulated with 1 μM CpG-ODN 48 h later, and NF-κB-dependent luciferase activity was determined as before (mean ± SD, one of three experiments).
An N-terminal positively charged area is important for TLR9 activity
In TLR3 two binding sites for RNA have been reported (8, 9) and in the N-terminal site conserved histidine residues H39 and H60 are involved in ribose backbone binding. This site is located in one of two positively charged areas in the TLR3 N terminus (Fig. 4⇓A). We therefore analyzed conserved histidine residues in the N-terminal part of TLR9 ECD, which are likely to carry a positive charge under the mild acidic pH conditions in the endosome and thus would be charge-complementary to the DNA backbone (17). Three conserved histidine residues, H76, H77, and H79, in LRR-NT and LRR1 of TLR9 were found to form a positive patch. These were mutated to phenylalanine or glutamic acid residues thereby removing positive or introducing negative charges, respectively. We also mutated two positively charged residues, K51 and R74, which generate a second charged patch in close proximity to the three histidines and adjacent to the irregular insert in LRR2 (Fig. 4⇓A). Surprisingly, mutations of the histidines had no effect on CpG-induced TLR9 activation (Fig. 4⇓B). In contrast, mutating K51 or R74 to the uncharged amino acid methionine led to nearly complete loss of function of the receptor. The glutamate mutant did not have stronger effects in the case of R74, whereas for K51 the residual activity still present in the K51M mutant vanished for K51E.
A positively charged patch around lysine 51 and arginine 74 is essential for TLR9 signaling. A, Surface charge computation of the murine TLR3 (left panel) and murine TLR9 (right panel) N termini. Red, negative charge; blue, positive charge. Distinct positively charged patches marked by dashed circles and individual residues of interest labeled. B and C, HEK293 cells were transfected with the indicated plasmids. B, TLR9 function was assessed upon stimulation with CpG-ODN by NF-κB reporter gene activity as before. C, TLR9 expression was analyzed by Western Blotting. D, Binding of biotinylated CpG-ODN to HA-tagged TLR9 was determined by streptavidin-agarose pull-down and quantitative Western blotting.
The N-terminal point mutations did not affect expression of TLR9 as all mutants were detected at comparable levels and at the expected size in Western blot analysis (Fig. 4⇑C). As shown before for all other mutants, a second band of lower m.w. occurred, indicating that cleavage of the receptor was not influenced by the mutants. However, the point mutants that lacked activity (K51M and R74M) showed abolished CpG-ODN binding, whereas the histidine mutants were still able to interact with CpG-ODN (Fig. 4⇑D). The results confirm that the identified N-terminal patch represents a ligand interaction site.
Decreased activity of TLR9 mutants is not due to impaired TLR9 cleavage
Recently, it has been published that the ECD of TLR9 is cleaved within the endosome to generate a functional receptor (15, 16). Furthermore, it was suggested that the C-terminal cleavage fragment (encompassing LRR15s and further) would be sufficient to mediate receptor activation. Although a different model system was used, such a mode of ligand recognition contrasts with our observation of a specific contribution of both N-terminal insertions (LRR2, 5, and 8) and a specific N-terminal positively charged patch (K51 and R74) to CpG-DNA recognition. Reexamining our protein expression data of C-terminally tagged TLR9-HA mutants we could observe a second band in Western blots of approximately the size published (90 kDa; cf. Figs. 2⇑D and 4⇑C). This band could be observed for all mutants analyzed, so that the loss of function of several of these mutants is unlikely to be due to interference with the cleavage reaction. Moreover, in the coimmunoprecipitation studies we performed, the putative C-terminal fragment was also present in the precipitate (Figs. 1⇑D and 2⇑D). Because several point mutations or deletions within the N-terminal part of TLR9 ECD abolished TLR function, our results argue for a role of this part or fragment in receptor activation, even though the ECD is cleaved and that the C-terminal fragment by itself is capable of CpG-ODN-binding (16).
Cathepsins B and L were shown to be able of cleaving TLR9 in vitro, but inhibitors did not influence TLR9 processing and mice deficient of different cathepsins had no defect in TLR9 signaling (15). Inhibition of lysosomal cystein proteases using the inhibitor z-FA-fmk inhibited TLR9 but not TLR4 signaling in a recent report (16), whereas a previous publication also observed inhibition of LPS signaling due to interference with NF-κB transactivation (29). We therefore reanalyzed the effects of z-FA-fmk on TLR signaling in RAW264.7 macrophages. We observed that z-FA-fmk did not only inhibit TLR7 and TLR9 signaling but also interfered with LPS-induced NO and TNF-α secretion from RAW264.7 cells (Fig. 5⇓, A and B), suggesting that it did not specifically inhibit endosomally located TLR. The effects were not due to cytotoxicity because MTT assays showed no alterations by inhibitor treatment (data not shown). To further exclude that the loss of function of K51M and R74M was due to effects on receptor cleavage, we introduced the same mutations into the chimeric receptor TLR9N4C (30). This receptor is composed of the ECD of TLR9 and the transmembrane and cytoplasmic domain of TLR4, thereby localizing to the cell surface. As reported by others (16), we could observe no cleavage product of TLR9N4C WT in Western blot analyses (Fig. 5⇓D). Nevertheless, point mutations in the ECD of TLR9N4C did affect receptor function in a similar way as observed for TLR9 WT: K51M and R74M led to complete loss of CpG-induced activation whereas H77F had no influence (Fig. 5⇓C). In Western blot analyses, the mutant forms of TLR9N4C showed only a single band, indicating that despite the lack of cleavage of TLR9N4C, the residues K51 and R74 are critical for receptor activation in this chimeric receptor. Taken together, the results rule out an interference with receptor cleavage to be causative for improper function of the mutants K51 and R74.
N-terminal recognition of CpG-DNA is independent of TLR9 cleavage. A and B, RAW264.7 cells were preincubated with 5, 10, or 20 μM z-FA-FMK for 2 h and then stimulated with 300 nM CpG-ODN, 10 ng/ml LPS, or 300 ng/ml R848. Twenty-four hours or 48 h later, TNF-α and NO2−, respectively, were determined in the supernatant. C and D, HEK293 cells were transfected with the indicated plasmids. TLR9 function was assessed upon stimulation with CpG-ODN by NF-κB reporter gene activity (C), and TLR9 expression was analyzed by western blotting (D) (∗, possible cleavage product of TLR9).
Discussion
To identify regions in TLR9 involved in receptor activation and ligand binding, we conducted a mutational approach with a focus on the role of irregular LRR, which contain inserting sequences. In TLR9, such insertions can be observed in LRR2, 5, 8, and 11 (inserting at position 10) and in LRR20 (after position 15). We first deleted these insertion-bearing LRR and one additional regular LRR (LRR16). Our data suggest that the correct number of LRR modules within the ECD is essential for proper receptor function, because deletion of any complete LRR tested led to complete loss of responsiveness toward CpG-ODN, irrespective of the position of the LRR and whether it was irregular or not. Taken together, the results of this first approach showed that the spatial arrangement of individual LRR in TLR9 is crucial for proper receptor function.
Interestingly, mutants with missing LRR were still able to form preformed dimers but lost the ability to bind CpG-ODN (data not shown). These data sharply contrast published data for TLR2 where it was possible to delete the seven N-terminal LRR without disturbing the recognition of Pam3CSK4 (31). Thus, in TLR2, the recognition site for lipopeptides was mapped to a distinct region by mutational studies, and these findings were subsequently confirmed by x-ray crystallography (5). In TLR3, deletion of some (but not all) complete LRR was also shown to render TLR3 unresponsive (32). As even the deletion or switching of typical LRR domains was not tolerated in TLR9, this receptor appears even more sensitive to such manipulation, and ligand recognition in TLR9 follows different spatial constraints.
However, by deleting only the inserted sequences in the irregular LRR, we identified specific LRR (LRR2, 5, and 8) in the N-terminal part of the ECD to be essential for receptor function. In contrast, the lack of the insertions in LRR11 and 20 did not or only partly interfere with TLR9 activity. Our binding data suggest that loss of function was due to decreased binding of the respective mutants to CpG-DNA, highlighting the importance of LRRs2, 5, and 8 for CpG-ODN engagement.
The inactive mutants were, as expected from the results before, still able to dimerize with TLR9 WT and acted in a dominant-negative manner on TLR9 WT stimulation. This suggests that in TLR9 C-terminal regions (e.g., C-terminal LRR and LRR-CT) engage in receptor-receptor interactions similar to TLR3 (9) or Drosophila Toll (33, 34), whereas ligand binding necessarily requires the insertions in N-terminal LRR. Our data do not rule out that the region located between LRR11 and 20 may harbor an additional nucleic acid binding site similar to the C-terminal binding site for dsRNA by TLR3 (8, 9) (Fig. 6⇓). Indeed, it has been reported that LRR17 contains a region with sequence similarity to methyl-CpG binding proteins (35). D535 and Y537 are located in this region and on the same side of the TLR9 ECD as K51, R74, and the LRR insertions according to our homology model (Fig. 6⇓, A and D). Mutation of these residues to alanine was shown to abrogate TLR9 responsiveness to and binding of CpG-ODN (36). We therefore suggest that the region surrounding D535 and Y537 may be the functional equivalent of the second binding site for dsRNA in TLR3 (residues N515, N517, H539, N541, and R544; Fig. 6⇓) (9).
Nucleic acid recognition in TLR3 and 9 follows similar structural principles. A, Surface representations of the generated TLR9 ECD model in side-view (top) and upon 90° rotation (horizontal: bottom left, vertical: bottom right). Proposed two binding regions are highlighted by boxes. Insertions of the LRR2, 5, and 8 contributing to receptor activation are depicted in green colors. Patch 2 residues in the N terminus, important for signaling, are shown in red, patch 1 residues, dispensable for signaling, are shown in blue. Putative DNA positioning is marked as pink, dashed region. B, Analogous recognition of dsRNA by TLR3. Surface representation of complex structure (pdb code: 3ciy) with two binding regions highlighted by boxes, poly(I:C) ligand in pink. C and D, Close-up of the proposed N- and C-terminal regions of TLR9 responsible for CpG-ODN binding (36 ). Residues identified as important for ligand binding and receptor activation are highlighted in red. Important residues are shown in ball-and-stick representation in red.
Furthermore, we could show that the lack of the insertions in LRR2, 5, and 8 led to diminished ligand binding. Thus, in TLR9, these specific insertions are involved in the recognition of ligand. To test whether the irregular LRR contribute to the ligand specificity of TLR9 and 7, we substituted the irregular LRR in TLR9 by the corresponding ones of TLR7. Unexpectedly, even the exchange of only one LRR led to loss of responsiveness toward the TLR9 ligand CpG-ODN. Surprisingly, this was also true for LRR16, which fits the consensus LRR motif perfectly. Moreover, we could not turn the specificity toward R848, even when all irregular LRR from TLR9 were shuffled to TLR7. This again contrasts with data from TLR2 and 6. In these TLR, the only LRR insertion is located in LRR11. In mutational studies, Meng et al. (31) switched regions between human and murine TLR2, showing that the species-specific response to trilauroylated lipopeptides was mediated by LRR7–10 of TLR2 and that segments of the TLR2 ECD can be modified without the loss of receptor function. In our experiments, exchange of even the most highly homologous LRR6 (or parts thereof) in TLR9 by that of TLR7 resulted in deleterious effects on TLR9 responses.
For TLR3, it has been reported that several conserved histidine residues in the N-terminal part of the ECD directly bind to the negatively charged ribose backbone of RNA. It was proposed that this contributes to the pH dependency of RNA binding to TLR3, because in the mildly acidified endosomal compartments, the histidine imidazole rings would be protonated and thus charge-complementary to the RNA backbone. Our results also suggest a role for the N-terminal part of the ECD in binding of the nucleic acid ligand of TLR9. However, whereas the N-terminal histidines 76, 77, and 79 located in LRR-NT and LRR1 were dispensable for function, we observed that two positively charged amino acids located in close proximity of the histidine residues, namely K51 and R74, were of crucial importance. Substitution of K51 or R74 with neutral amino acids led to complete inactivity or in the case of R74M to a strongly diminished response of TLR9. By introducing a negative charge at position 74, this residual activity was abolished. It is interesting to note that TLR3 also contains two positively charged patches (Fig. 4⇑A). The first one is dispensable for signaling as illustrated by the fact that R64 and R65 can be mutated to glutamic acid without loss of function (8). The second one, however, containing H39 and H60, is involved and ligand binding and mutation to alanine or lysine abrogates signaling (8, 9). Our data suggest that K51 and R74 are the functional equivalent of this TLR3 histidine patch and that the histidines in TLR9 do not play a role in ligand binding, similarly to R64 and R65 in TLR3. This was confirmed by ligand-binding studies showing that these two amino acids were involved in binding of CpG-ODN, thereby showing that surface charges in patch 2 but not patch 1 contribute to receptor activation in TLR9. Structural determinants of ligand binding may thus be similar in the nucleic acid sensors TLR3 and TLR9 but may depend on different residues.
Recently, it was shown that the ECD of TLR9 undergoes cleavage in the endosome, and it was proposed that this step is indispensable for TLR9 function (15, 16). Such a mode would imply that the C-terminal cleavage fragment is sufficient for CpG-DNA binding and receptor activation and would hardly be compatible with our findings that clearly indicate a role of the N terminus of TLR9 for receptor activation. We can show that the putative cleavage product can be observed for all mutants (but not for the TLR9 constructs located to the cell surface), thus ruling out that the mutants show loss of activity due to failure to undergo cleavage. It is possible that these seemingly contrasting sets of data could be reconciled if C-terminal fragment and full-length receptor chains remain strongly associated after cleavage—a feature that was recently shown for Drosophila Toll (33)—and that this complex has a higher affinity for CpG-DNA. To clarify this point, it would be of interest to compare the reported immunoprecipitation results with binding studies using CpG-ODN and purified TLR9 ectodomain (36), as well as the truncations and point mutations described here and elsewhere (15, 16). Moreover, in our hands, the cathepsinB/L inhibitor z-FA-fmk did not specifically inhibit signaling from endosomally located TLR but affected NF-κB activation in response to multiple TLR as observed previously (29).
We conclude that the N-terminal part of the ECD of TLR9 is necessary for proper CpG-DNA-mediated receptor activation as multiple mutations in this region showed loss of function without affecting expression, dimerization, or cleavage. We do not exclude that cleavage of TLR9 occurs at some stage of receptor activation and that a C-terminal fragment may play a distinct role in signal activation or regulation, but our data argue that the N terminus is necessary for full receptor activation in the HEK293 complementation system. More specifically we suggest that the positively charged patch 2 in LRR-NT and LRR1, which also is in proximity to the insertions of LRR2 and 5 (Fig. 6⇑C), represents a binding site for CpG-DNA similar to such a site in TLR3.
In light of the previously reported findings on D535 and Y537, we thus propose a model for nucleic acid recognition by TLR9 (Fig. 6⇑A) similar to the ones put forward recently for TLR3 (8, 9) (Fig. 6⇑B), in which the nucleic acid ligand is bound at two distinct receptor sites. Although it is difficult to envisage how such a mode of binding could be used by small-molecule agonists such as R848, we propose that nucleic acid recognition by TLR7 and 8 may follow a similar principle. The additional ligand interactions mediated by the TLR9 LRR insertions might induce steric forces that require more flexibility, and this could be achieved by the low-consensus region between LRR14 and 15 whose structure is unknown, finally resulting in a decrease of the ECD diameter upon CpG-specific binding to TLR9 (11). In our model (Fig. 6⇑A), the very C-terminal part of the ECD of TLR9 would be responsible for protein-protein interactions resulting in the preformed dimer. This would explain why none of the mutants analyzed (all are located between LRR-NT and LRR20) influenced dimer formation. This is in concordance with the m-shaped dimers observed in the crystal structures of other TLR ligand-receptor complexes, which show individual ECD converging in the middle with their C termini (5, 9). The conformational change induced in each ECD by engagement of the ligand at N- and C-terminal parts of the ECD could be relayed to the Toll/IL-1 receptor domains via this fixed point of contact, leading to a new relative orientation of Toll/IL-1 receptor domains that enables cytoplasmic signaling. The requirement for an exact positioning of N- and C-terminal ligand binding patches, LRR insertions and receptor-receptor dimerization regions may explain why TLR9 is so sensitive to subtle changes in the LRR horseshoe compared with other TLR. Further structural and functional studies experiments outside the scope of this present work will be required to unravel the precise determinants of ligand recognition by TLR9 and other nucleic acid detecting receptors.
Acknowledgments
We thank J. George for helpful discussions and M. Frank and W. von der Lieth for computer and modeling assistance. We acknowledge technical support from C. Schütz and A. Dillmann.
Disclosures
The authors have no financial conflict of interest.
Footnotes
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 This study was supported by the German Research Foundation (Deutsche Forschungsgemeinschaft DA592/3 (to A.H.D.); Emmy Noether program (to A.N.R.W.)) and by the German Cancer Research Center (to A.V.K., A.N.R.W.).
↵2 M.E.P. and A.V.K. contributed equally.
↵3 Address correspondence and reprint requests to Dr. Alexander H. Dalpke, Department of Medical Microbiology and Hygiene, Hygiene-Institute, University Heidelberg, Im Neuenheimer Feld 324, 69120 Heidelberg, Germany; E-mail address: alexander.dalpke{at}med.uni-heidelberg.de or Dr. Alexander N. R. Weber, Junior Group “Toll-like Receptors and Cancer,” German Cancer Research Center, Im Neuenheimer Feld 242, 69120 Heidelberg, Germany. E-mail address: alexander.weber{at}dkfz.de
↵4 Abbreviations used in this paper: ECD, extracellular domain; HA, hemagglutinin; LRR, leucine-rich repeat; ODN, oligodeoxynucleotide; WT, wild type.
- Received March 12, 2009.
- Accepted April 10, 2009.
- Copyright © 2009 by The American Association of Immunologists, Inc.
References
In this issue
