Cytoplasmic Targeting Motifs Control Localization of Toll-like Receptor 9*

Toll-like receptors (TLRs) are essential for host defense. Although several TLRs reside on the cell surface, nucleic acid recognition of TLRs occurs intracellularly. For example, the receptor for CpG containing bacterial and viral DNA, TLR9, is retained in the endoplasmic reticulum. Recent evidence suggests that the localization of TLR9 is critical for appropriate ligand recognition. Here we have defined which structural features of the TLR9 molecule control its intracellular localization. Both the cytoplasmic and ectodomains of TLR9 contain sufficient information, whereas the transmembrane domain plays no role in intracellular localization. We identify a 14-amino acid stretch that directs TLR9 intracellularly and confers intracellular localization to the normally cell surface-expressed TLR4. Truncation or mutation of the cytoplasmic tail of TLR9 reveals a vesicle localization motif that targets early endosomes. We propose a model whereby modification of the cytoplasmic tail of TLR9 results in trafficking to early endosomes where it encounters CpG DNA.

Innate recognition of pathogens is mediated by a number of germ line-encoded proteins, of which the Toll-like receptors (TLRs) 3 play a pivotal role. TLRs comprise a family of approximately 10 paralogs, depending on species, which initiate inflammatory responses to a diverse array of microbial products (1,2). Examples include bacterial peptidoglycans, lipopolysaccharide (LPS), and flagellin, which are recognized by TLRs 2, 4, and 5, respectively, and nucleic acids from bacteria and viruses, which are recognized by TLRs 3,7,8,and 9 (1,3,4). TLR9, the focus of the current study, responds to bacterial and viral DNA as well as synthetic oligodeoxynucleotides that contain unmethylated CpG dinucleotides in specific sequence con-texts (CpG DNA) (5)(6)(7)(8). Like all TLRs, TLR9 is a type 1 integral membrane receptor with a ligand-binding ectodomain and a cytoplasmic portion that belongs to the Toll-IL-1 receptor (TIR) family of signaling domains. The TIR domain of TLR9 interacts with the MyD88 adapter protein, initiating a signaling cascade that results in the activation of NF-B and mitogenactivated protein kinases (MAPKs) (9,10).
TLRs recognize their ligands at distinct locations within cells. LPS and peptidoglycan are recognized at the cell surface (11)(12)(13); however, recognition of flagellin by TLR5 is more compartmentalized and occurs on the basolateral surface of polarized epithelial cells (14). In contrast, nucleic acid recognition occurs intracellularly (13,(15)(16)(17). TLR9 is retained in the endoplasmic reticulum (ER) prior to CpG DNA exposure (18,19), and it has been proposed that TLR9, localized in the ER, gains access to incoming endosomes containing CpG DNA by direct fusion of the endosome with the ER. Neutralization of lysosomes inhibits signaling through TLR9, suggesting that trafficking to the lysosome is essential for signal transduction (13,15). Furthermore, TLR9, CpG DNA, and the adapter protein MyD88 have all been localized to vesicles that can be stained for lysosomal markers (5,13). The endosomal location of TLR9 reflects the requirement for phagocytosis and microbial destruction for the release of DNA prior to signaling.
In the current study, we hypothesized that TLR9 contains specific motifs that control localization and trafficking. Although transmembrane proteins often contain information in their cytoplasmic tails that dictate their cellular localization (20,21), we found that the internal location of TLR9 was mediated by both the cytoplasmic domain and the ectodomain, but not the transmembrane domain. Truncation analysis revealed a region of the cytoplasmic tail required for intracellular retention and a motif that was capable of targeting TLR9 to vesicles. Unexpectedly, the vesicles targeted by this motif were not lysosomes but instead were early endosomes coincident with endocytosed CpG DNA.

TLR9 Localization Motifs
directly labeled with Alexa Fluor 488 using the Molecular Probes monoclonal antibody labeling kit. Anti-IL-2R␣ (N-19), used for immunoblotting Tac chimeras, was from Santa Cruz Biotechnology. Unconjugated anti-TLR4 (HTA125) and TLR9 (eB72-1665) were from eBioscience. LAMP-1 antibody was from Pharmingen, allophycocyaninconjugated secondary antibodies were from Caltag, and horseradish peroxidase-conjugated secondary antibodies were from Southern Biotechnology. Plasmids-TLR9 and TLR4 (GenBank TM accession numbers AF259262 and U93091, respectively), tagged at their C termini with enhanced green fluorescent protein (GFP), were generated by PCR and inserted in-frame into pEGFP-N1 (Clontech). Constructs containing the extracellular domain of Tac fused to the transmembrane (TM) and TIR domains of TLR9 (residues 818 -1032) and TLR4 (residues 635-835) were generated by introducing an EcoRV site directly upstream of the TM domain of the TLR by site-directed mutagenesis. The TM, TIR domain, and stop codon were excised at the EcoRV and a downstream restriction site and then ligated in-frame into compatible sites of pCDNA3 containing the extracellular domain of Tac (residues 1-239) (GenBank TM accession number NM_000417; a gift from Dr. Remy Bosselut, NCI). The TLR4-9-GFP (TLR4 1-633 TLR9 818 -1032 ) and TLR9-4-GFP (TLR9 1-817TLR4 634 -835 ) chimeras were generated by inserting the TLR9 (or TLR4) TIR and TM domains into TLR4 (or TLR9)-GFP from which the TIR and TM domains had been excised. TLR9 1-810 TLR4 625-835 was generated by PCR as described previously (26) and put into pFlag-CMV. Truncation mutants were generated by introducing stop codons into the Tac-TIR9 chimera by site-directed mutagenesis using the QuikChange mutagenesis kit (Stratagene) according to the provider's protocol except that the PCR reactions contained 200 ng of template plasmid and were run for 28 cycles (see Fig. 2). TLR4-9-4 (TLR4 1-678 TLR9 875-911 TLR4 710 -839 ) was generated by stitching PCR. All chimeras were verified by sequencing. Rab5-GFP was a gift from Dr. Juan Bonifacino, NICHD (National Institutes of Health). All plasmids for transfection were prepared using an EndoFree Plasmid Maxi Kit (Qiagen).
Flow Cytometry-HEK293T (3 ϫ 10 5 ) cells were transfected using TransIT (Mirus) in 6-well plates and after 24 h were stained directly for surface expression or fixed, permeabilized, and stained for total expression. Data were collected with a FACSCalibur (BD Biosciences) and analyzed using CellQuest.
Confocal Microscopy-HeLa (3 ϫ 10 5 ) cells were transfected using TransIT on coverslips and then fixed, permeabilized, and stained 24 h later. Cells were visualized using either a Zeiss LSM 510 confocal laser-scanning microscope, where images of single 0.8-m sections were collected, or by a Zeiss AxioImager for non-confocal images. Composite pictures were prepared with Photoshop. Each individual cell image is a representative of at least five different cells collected in at least three different experiments. Larger field images were taken with a ϫ40 objective from at least three independent experiments.
Luciferase Assay-HEK293 cells were plated at 2 ϫ 10 4 cells/ well in 96-well plates. Cells were transfected using TransIT and a total of 200 ng DNA/well consisting of NF-B-luciferase and pSV-␤-galactosidase reporter plasmids and either TLR9, TLR9-4 chimera, MD-2 plus TLR4, or MD-2 plus the TLR4-9 chimera. Cells were stimulated for 24 h, lysed in reporter lysis buffer (Promega), and assayed for luciferase (Promega) and ␤-galactosidase (Tropix) activity. NF-B activity (relative light units) was calculated by dividing the luciferase counts by the control ␤-galactosidase. Salmonella LPS (Sigma) or CpG oligodeoxynucleotide 2006 (23) were used as stimulants. Paired t tests were used to determine statistical significance.
Glycosidase Treatment-Transfected HeLa cells were collected in phosphate-buffered saline containing 20 mM EDTA with protease inhibitors and sonicated. Sonicates were treated with endoglycosidase H (Endo-H) or peptide N-glycosidase F (PNGase-F), both from New England Biolabs, as described (24). FIGURE 1. Both the ecto-and TIR domains of TLR9 localize intracellularly. A, HEK293T cells were transfected with TLR4-GFP, TLR9-GFP, TLR4-9-GFP, TLR9-4-GFP, or TLR9⌬TIR-GFP (TLR9-GFP with the TIR domain deleted). After 24 h, the cells were stained directly (Surface) or fixed, permeabilized, and stained (Total) with anti-TLR4 or anti-TLR9 monoclonal antibody (black lines). Cells transfected with GFP and stained identically served as negative controls (filled histograms). Cells were gated for GFP expression, and only the GFP ϩ cells are shown. Data are representative of three separate experiments. B, HeLa cells were transfected with the same constructs as described in A and fixed 24 h later. GFP fluorescence and differential interference contrast overlays are shown. C, HEK293 cells were transfected with TLR9-4 or TLR9 alone, with TLR4 plus MD2, or with TLR4-9 plus MD2 and stimulated with 100 ng/ml LPS or 5 g/ml CpG DNA. NF-B activity was measured 24 h later. *, p Ͻ 0.002; #, p Ͻ 0.05. This is one of three replicate experiments.  NOVEMBER 17, 2006 • VOLUME 281 • NUMBER 46

JOURNAL OF BIOLOGICAL CHEMISTRY 35587
Alternatively, cells were lysed as described previously (18), immunoprecipitated, and treated with Endo-H (New England Biolabs). All treated samples were separated by SDS-PAGE, transferred to nitrocellulose, and probed with anti-Tac or anti-GFP antibodies.

RESULTS
The localization of nucleic acid recognizing TLRs is critical for ligand recognition. To determine what molecular mechanisms control TLR9 intracellular localization, we constructed chimeric proteins in which the ecto-and transmembrane/cytoplasmic (TM-TIR) domains of TLR4 and TLR9 were interchanged. As expected, TLR4, but not TLR9, localized to the cell surface (Fig. 1A). However, no surface expression was detected for TLR9 or the TLR4-9 or TLR9-4 chimeras, suggesting that both the ecto-and TM-TIR domains of TLR9 contain intracellular localization information. A chimera in which the TIR domain of TLR9 was replaced by GFP (TLR9⌬TIR) also failed to localize to the cell surface (Fig. 1A), further supporting the notion that the TM domain and/or the ectodomain of TLR9 contains localization information. Confocal laser scanning microscopy revealed that TLR9-4, TLR4-9, and TLR9⌬TIR, like wild type TLR9, were retained in a reticular compartment, which we previously described to be the ER (Fig. 1B) (19). As expected, TLR4 was found both at the cell surface and in a Golgi-like compartment (Fig. 1B) (12,18,19). The TLR4-9 chimera failed to respond to LPS, likely because of redirection of the TLR4 ectodomain from the cell surface to intracellular compartments where it could no longer interact with LPS (Fig. 1C). By contrast, the TLR9-4 chimera, containing the TLR4 TM and cytoplasmic tail, did respond to CpG DNA. In this study we focused on molecular structures contained in the cytoplasmic tail that dictate intracellular localization.
The studies shown in Fig. 1 suggested that either the TM or the cytoplasmic domain, in addition to the ectodomain, contained intracellular localization motifs. Other studies have suggested that the TM domain controls localization of TLR7 and mouse TLR9 (25)(26)(27). To determine whether the TM domain of human TLR9 also controls intracellular localization, the ectodomain of CD25 (Tac) was fused to the TM and cytoplasmic tail of TLR9 (Tac-TIR9) (Fig. 2), a strategy that has been used successfully in other systems (24). Tac is processed in the Golgi to a highly glycosylated form that is resistant to Endo-H digestion, serving as a sensitive biochemical marker for protein export from the ER. A fusion of Tac and the TLR4 TIR domain (Tac-TIR4), readily detected on the cell surface by flow cytometry (supplemental Fig. 1), contained a high molecular weight Endo-H-resistant band indicative of Golgi processing. Tac-TIR9, which was localized intracellularly and co-localized with wild type TLR9, was completely Endo-H-sensitive (supplemental Fig. 1). In contrast, truncation of most or all of the cytoplasmic tail resulted in surface expression of Tac-TIR9 ( Fig. 3A and  supplemental Fig. 2). The Tac-TIR9 truncations contained higher molecular weight, Endo-H-resistant glycoforms indicative of trafficking out of the ER and of Golgi processing (Fig. 3B). Because those constructs contained the TLR9 TM domain, we concluded that the human TLR9 transmembrane domain did not mediate intracellular localization. In a previous report, surface expression of mouse FLAG-TLR9 1-810 TLR4 625-835 was detected (26). We generated the same mouse TLR9 chimera along with the analogous untagged human version and detected no mature glycoform, which would indicate surface expression, in HeLa or MEF cells (supplemental Figs. 3 and 4).
To identify motifs that control localization of TLR9, the Tac-TIR9 fusion protein was truncated progressively from the C terminus (Fig. 2). The cytoplasmic domain of TLR9 contains 196 residues, and removal of the last 131 amino acids (i.e. truncation at amino acid 902, TacTIR9 ⌬902 ) had little effect on localization, as this mutant was not detected on the cell surface and remained Endo-H-sensitive (Fig. 4). However, an abrupt transition from internal to surface expression was observed when 14 additional residues were deleted (TacTIR9 ⌬888 ) as detected by microscopy (Fig. 4A), flow cytometry (supplemental Fig. 5), and glycoform analysis (Fig. 4B). The 14 residues between TacTIR9 ⌬902 and TacTIR9 ⌬888 represented a functional intracellular localization motif because they directed internal localization when grafted into the homologous position in the TLR4 cytoplamsic domain (Tac-TIR4-9-4; Fig. 4 and supplemental Fig. 5). Although there was still surface expression of Tac-TIR4-9-4, the level was significantly reduced compared with Tac-TIR4 (mean fluorescence intensity (MFI) ϭ 98 versus MFI ϭ 230, respectively). A chimera in which the 14 residues were grafted into the TIR domain of full-length TLR4 (TLR4-9-4) had less cell surface expression (data not shown) and contained much less of the Endo-H-resistant glycoform (Fig. 5A) compared with wild type TLR4. TLR4-9-4 also lacked response to LPS; however, both TLR4-9 and TLR4-9-4 demonstrated increased constitutive activity when overexpressed, suggesting that the cytoplasmic tail retained the ability to transduce signals (Fig. 5B). Together these data indicated that the 14 amino acids between residues 888 and 902 of the TLR9 TIR domain mediated intracellular localization. It is important to note that these results do not rule out other intracellular localization motifs C-terminal to amino acid 902. In fact, a more detailed analysis of cytoplasmic tail truncation mutants revealed several changes in the intracellular staining pattern (Fig. 6A) and glycosylation pattern (Fig. 6B), suggesting there were several, possibly interrelated, localization motifs. Alexa488-labeled anti-Tac or fixed, permeabilized, and stained. Images were taken on a Zeiss AxioImager at the same settings and compiled in Photoshop. B, HeLa cells were transfected in 6-well plates and 24 h later were harvested for glycosylation analysis as described in the legend for Fig. 3. The top and bottom panels were from two separate gels developed simultaneously and are shown at the same exposure. The TacTIR9 ⌬902 samples were separated by the molecular weight marker in the original gel. NOVEMBER 17, 2006 • VOLUME 281 • NUMBER 46

JOURNAL OF BIOLOGICAL CHEMISTRY 35589
TacTIR9 ⌬888 represented a transition from mostly intracellular to almost exclusively cell surface expression. Because several truncations downstream of amino acid 888 displayed some vesicular staining (Fig. 6), we asked which vesicular compartment was targeted by our mutant and whether that compartment was coincident with CpG DNA following its uptake from the medium. TacTIR9 ⌬898 showed significant targeting to endosomes and some cell surface expression (Fig. 7). Furthermore, co-localization in vesicles with fluorescently labeled CpG DNA was detected when HeLa cells were transfected with TacTIR9 ⌬898 (Fig. 7A). Unexpectedly, these vesicles were devoid of LAMP-1 (Fig.  7B) but instead contained an early endosome marker, Rab5 (Fig. 7C). Although some studies have indicated that TLR9 co-localizes with CpG DNA in lysosomes (13), our data are consistent with studies showing that intact TLR9 co-localizes with CpG DNA in early endosomes prior to reaching lysosomes (18). We predicted that there was a motif that targets vesicle localization in the region surrounding amino acid 888. Inspection of the sequence in this region revealed a potential tyrosine-based vesicle localization motif, YXX⌽ (residues 888 -893, where ⌽ is a hydrophobic amino acid) (21). Truncation mutants containing the YXX⌽ motif were localized intracellularly and in vesicles (TacTIR9 ⌬898 , Tac-TIR9 ⌬902 ). However, truncation mutants lacking the YXX⌽ motif were redistributed to the cell surface, supporting our hypothesis that the cytoplasmic tail of TLR9 contains a tyrosine-based vesicle-targeting motif.

DISCUSSION
Bacterial DNA, a potent stimulus of TLR9, is contained within microorganisms and becomes accessible for recognition only after lysis in phagolysosomes. It has been shown that TLR9 co-localizes with CpG DNA in cytoplasmic vesicles but resides in the ER prior to exposure to CpG DNA (5,13,19). Moreover, it has been proposed that TLR9 reaches the endocytic pathway through direct fusion of the ER with the CpG DNA containing endosome (18). What remains FIGURE 5. 14-amino acid region confers intracellular localization to TLR4. A, HeLa cells were transfected in 6-well plates with the indicated GFP-tagged chimeras and 24 h later were harvested for GFP immunoprecipitation and glycosylation analysis as shown in Fig. 3. The second and third lanes were separated by the molecular weight marker in the original gel. B, HEK293 cells were transfected with MD2 plus 1.5 or 150 ng/well TLR4, TLR4-9, or TLR4-9-4. Cells were stimulated with 100 ng/ml LPS, and NF-B activity was measured 24 h later. unknown is how the localization and movement of TLR9 is orchestrated. Here we show that both the ecto-and TIR domains of TLR9 contain sufficient information for intracellular localization and trafficking, and we have identified a specific region within the TIR domain that contributes to intracellular localization and to vesicle targeting. By creating C-terminal truncation mutants in the Tac-TIR9 chimera we identified a 14-amino acid stretch between residues tyrosine 888 and tryptophan 902 (Fig. 2) responsible for intracellular localization. The region contains a tyrosine-based vesicletargeting motif, YNEL. Truncation of this motif leads to a loss in vesicular localization and redistribution to the cell surface.
Recent data from other laboratories support the idea that the transmembrane domain of mouse TLR9 controls intracellular localization (26,27). We have been unable to reproduce these data. Neither FLAG-tagged mouse nor the untagged human TLR9-4 chimera (TLR9 1-817 TLR4 634 -835 ) was detected on the cell surface in HEK293, HEK293T, MEF, or HeLa cells (supplemental Figs. 3 and 4 and data not shown). We failed to detect surface expression or mature glycoforms even if the chimera junction, like the previously described chimera, was seven amino acids upstream of the TM domain (TLR9 1-810 TLR4 625-835 ). Furthermore, we gen-erated a soluble TLR9 ectodomain and found that it was also retained in the ER, 4 suggesting that sequences in the ectodomain associate with chaperones or other luminal proteins to effect TLR9 ER localization. In our hands, a chimera containing the TM and cytoplasmic tail of TLR9 was localized like wild type. However, when most or all of the cytoplasmic tail was truncated from this mutant, the chimera was redistributed to the cell surface (Fig. 3) ruling out any role for the transmembrane domain in TLR9 localization. Finally, the region that we identified in the cytoplasmic tail could confer, at least partially, intracellular retention to the normally surface-expressed TLR4 in the context of either the Tac fusion or the full-length wild type TLR4. The redistribution of the TLR4-9-4 chimera to a mostly intracellular form eliminated the ability of TLR4 to recognize LPS.
The three-dimensional structure of the TLR9 TIR domain can be inferred by homology with the known structures of the TLR1, TLR2, and IL-1 receptor-associated protein TIR domains (28,29). Fig. 8 highlights the 888 -902 region (blue) on the predicted TIR domain structure of TLR9. This region lies on the surface in the first ␣-helix and the loop that connects the first helix with the second ␤-strand. Contiguous with this region is the BB loop, which contains a proline residue that is essential for signal transduction in most TLRs. In TLR2 this conserved proline is required for the binding of the TLR2-TIR domain to MyD88 (28). However, the proline is dispensable for TLR4 binding to MyD88 and MAL (30). If the TLR9 BB loop interacts with MyD88, as in TLR2, then the recruitment of MyD88 to TLR9 might influence the intracellular localization of TLR9 by interfering with its ability to interact with cytoplasmic localization adapter molecules. If the BB loop is not involved in TLR9-TIR binding to MyD88, as in TLR4, other adapter protein binding or receptor oligomerization may change the structure of the cytoplasmic tail in the region containing the BB loop and thus influence the ability of cytoplasmic localization adapter molecules to bind.
Although TLR9 is in the ER prior to stimulation and found in lysosomes with CpG DNA at later time points, how TLR9 traffics is not clear. One mechanism by which this might occur is the direct fusion of TLR9-containing ER with early endosomes 4 C. A. Leifer, personal observation. . Vesicle motif targets TIR9 to CpG DNA in early endosomes. A, HeLa cells were transfected with TacTIR9 ⌬898 and incubated with Cy-3-labeled CpG DNA for 1 h prior to fixation and staining. Tac is shown in green and CpG DNA in red. B, TacTIR9 ⌬898 -transfected HeLa cells were stained for Tac (green) and LAMP-1 (red). C, HeLa cells were co-transfected with TacTIR9 ⌬898 and Rab5-GFP (green), an early endosome marker, and stained for Tac (red). Single-slice images (0.8 M) were collected on a Zeiss LSM 510 Meta confocal microscope with a ϫ63 objective. Sections of overlays were magnified for better visualization of co-localization (insets). that contain CpG DNA (18). Alternatively, a CpG DNA-dependent signal upstream may induce the trafficking of TLR9 to the endosomal compartment. Previous results suggest that the TLR9 TIR domain may be tyrosine-phosphorylated in response to CpG stimulation (31), but we have been unable to detect phosphorylation of tyrosine 888. 4 Tyrosine 888 of TLR9 is not conserved in other intracellularly localized TLRs (TLRs 3, 7, and 8). Therefore, it is not clear whether tyrosine 888 plays a structural role in the formation of the localization site or serves a recognition function specific for TLR9. Although the surface receptor that would induce TLR-independent CpG DNA signaling events is unknown, surface recognition of CpG DNA by plasmacytoid dendritic cells has recently been shown to involve CXCL16 (32).
In summary, our data support a model in which a tyrosinebased localization motif in the cytoplasmic tail of TLR9 is involved in trafficking of TLR9 to early endosomes where CpG DNA signaling initiates through TLR9. Further studies are needed to identify which cellular proteins interact with the tyrosine motif and control TLR9 localization and trafficking. How this trafficking relates to the release of DNA following uptake of pathogens remains to be determined.