TLR9 acts as a first-line host defense against pathogens recognizing DNA comprising unmethylated CpG motifs present in bacteria and viruses. Species- and sequence-specific recognition differences were demonstrated for TLR9 receptors. Activation of human (h)TLR9 requires a pair of closely positioned CpG motifs within oligodeoxyribonucleotides (ODNs), whereas mouse TLR9 is effectively activated by an ODN with a single CpG motif. Molecular model-directed mutagenesis identified two regions, site A and site B, as important for receptor activation. Amino acid residues Gln346 and Arg348 within site A contribute to the sequence-specific recognition by hTLR9 in determining the bias for two appropriately spaced CpG motifs within immunostimulatory ODNs. Mutation of Gln562 at site B, in combination with Gln346 and Arg348 mutations of mouse counterparts, increased activation of hTLR9 by mouse-specific ODN, mammalian genomic DNA, and bacterial DNA. We propose that the double CpG motif sequence-specificity of hTLR9 results in decreased activation by ODNs with a lower frequency of CpG motifs, such as from mammalian genomic DNA, which increases hTLR9 selectivity for pathogen versus host DNA.

An intrinsic host defense against invading pathogens relies on the evolutionarily conserved innate immune response that consists of germline-encoded pattern recognition receptors, notably TLRs. Based on their cellular location, TLRs are classified into those residing in the plasma membrane, as well as a subfamily of nucleic acid–sensing receptors (TLR3, TLR7, TLR8, TLR9, as well as TLR13 found only in mice) whose recognition of nucleic acids is primarily confined to endosomes (1, 2). Activated TLR9 conveys a signal via the MyD88-dependent signaling cascade that culminates in the activation of NF-κB, AP1, and IRF7 transcription factors (1, 3). The inflammatory cytokines and chemokines instruct the development of Ag-specific adaptive immunity and play a key role in the recruitment of immune cells to the site of infection (1, 4).

The molecular structure of TLR9 was determined for horse TLR9, mouse (m)TLR9, and bovine TLR9 ectodomains (ECDs), in complex with a 12-nt stimulatory ODN1668, which is, however, a very weak TLR9 agonist, or with inhibitory oligodeoxyribonucleotides (ODNs). Two sites on horse TLR9 interact with each molecule of synthetic ODN1668_12nt in a 2:2 complex (5). The crystal structures of TLR9 and TLR8 ECDs, in complex with synthetic ligands, revealed that, despite proteolytic cleavage of the Z-loop, the two segments of the ECD remain stabilized by noncovalent interactions, essentially without structural consequences, and both are required for ligand recognition and receptor activation (57).

The natural source of CpG-containing ssDNA is bacterial or viral DNA, which is processed by endosomal DNase II to generate ssDNA fragments (8). Because DNA is also present in host cells, the receptor response has to be protected from activation by endogenous DNA by the selective recognition motif and by the methylation of host DNA, which are critical factors that limit TLR9 activation. Additionally, the higher frequency of inhibitory motifs in the mammalian genome contributes to the impaired activation by endogenous DNA (9). However, host DNA internalized by immune complexes can activate an immune response and lead to the development of autoimmune diseases (1, 1012). DNA released by apoptotic cells may also be a strong inducer of sterile inflammation (e.g., acetaminophen-induced hepatotoxicity) (13). The synthetic single-stranded ODN containing stimulatory CG motifs recapitulates activation of TLR9 with microbial DNA. B-class ODNs show sequence preference for activation of TLR9. The optimal hexamers within B-class ODNs are GACGTT (14, 15) and AACGTT (16) for activation of mTLR9 and GTCGTT for human and primates (17, 18) and other vertebrate species (19). Although mTLR9 can be activated by an ODN with >23 nt comprising a single CpG motif (20), human (h)TLR9 is activated by ODNs comprising at least two CpGs separated by 6–10 nucleotides, where the first CpG motif is preceded by the 5′-thymidine and having a length ≥20 nt (21).

Evaluation of the minimal ODN sequence motifs more precisely elucidated several previously observed differences in the ODN-binding motifs between mTLR9 and hTLR9. Aiming to unravel the molecular mechanism of those differences for ligand specificity, a structural model of hTLR9 ECD was constructed based on the recently determined structures of orthologous TLR9s (5). This hTLR9 model was used to guide the mutagenesis to investigate the residues responsible for the differences, as well as to identify a potential second CpG-recognition site of hTLR9. Residues within the two sites, A and B, positioned N- and C-terminal from the cleaved Z-loop and corresponding to the ligand binding sites of TLR8 (6), were identified as important for hTLR9 activity. Amino acid residues Q346 and R348, as part of site A, were singled out because their mutagenesis to mouse counterparts of hTLR9 abolished the stringent requirement for a two-CpG motif. Those differences were demonstrated on B cells and dendritic cells (DCs) transfected with wild-type (wt) and mutant TLR9. Moreover, the additional mutation Q562K within the proposed B site of hTLR9 improved activation efficiency. The relevance of the requirement for a pair of CpGs is demonstrated by the decreased activation of hTLR9 in comparison with mutations with mouse residues at proposed site A by genomic DNA (gDNA). This difference is particularly exhibited by the increased differentiation of hTLR9 between the gDNA of bacteria and host.

Human embryonic kidney cell lines HEK293 and HEK293T and mouse leukemic monocyte macrophages (RAW-Blue cells, which stably expressed an NF-κB/AP-1–inducible secreted embryonic alkaline phosphate [SEAP] reporter gene), derived from RAW264.7 (InvivoGen), were cultured in complete medium (DMEM; 1 g/l glucose, 2 mM l-glutamine, 10% heat-inactivated FBS; Life Technologies, Invitrogen) in 5% CO2 at 37°C. B lymphocytes (Ramos-Blue cells; InvivoGen), which stably expressed an NF-κB/AP-1–inducible SEAP reporter gene, were cultured in IMDM (Life Technologies, Invitrogen) supplemented with 10% (v/v) heat-inactivated FBS at 37°C in 5% CO2.

Plat-E cells (University of Tokyo) were cultured in DMEM (Life Technologies, Invitrogen) supplemented with 10% FBS, blasticidin (10 μg/ml), puromycin (1 μg/ml), penicillin G (100 U/ml), streptomycin (100 μg/ml), l-glutamine (0.3 mg/ml), and 2-ME (50 μM) in 5% CO2 at 37°C.

The murine pro-B cell line Ba/F3 expressing NF-κB–GFP was established as described (22). Cells were cultured in RPMI 1640 supplemented with 10% FBS, penicillin G (100 U/ml), streptomycin (100 μg/ml), l-glutamine (0.3 mg/ml), 2-ME (50 μM), and IL-3 (produced in CHO cells) in 5% CO2 at 37°C.

wt C57BL/6N mice were purchased from Japan SLC. TLR9−/− mice (C57BL/6N background) were generated by Dr. K. Miyake. Eight- to twelve-week-old age- and sex-matched male mice were used for experiments. Mice were kept in sterile pathogen-free conditions and used within the guidelines of the University of Tokyo.

Expression plasmids containing the hTLR9 gene (pUNO-hTLR9HA) and the mTLR9 gene (pUNO-mTLR9HA) were from InvivoGen; TLR9-YFP (pcDNA3-TLR9-YFP) was from Addgene. Expression plasmids containing sequences of hTLR9 (pUNO-hTLR9HA) were from InvivoGen. Plasmid phRL-TK constitutively expressing Renilla luciferase for normalization of transfection efficiency was from Promega. Plasmid coding for firefly luciferase under the NF-κB promoter (pELAM-1–luciferase) was a gift from C. Kirschning (Institute for Medical Microbiology, University of Duisburg-Essen, Essen, Germany), and plasmid-expressing mCerulean was provided by D. Piston (Vanderbilt University, Nashville, TN). For site-directed mutagenesis, C-terminal hemagglutinin (HA)-tagged hTLR9 (hTLR9HA), mTLR9 (mTLR9HA), mECD (m26–818 aa-hTLR9818–1032 aa-HA), and m504-hTLR9 (m26–504 aa-hTLR9504–1032 aa-HA) inserted into the pUNO vector were used. Expression plasmids pMX–IRES–rat CD2 (rCD2) and pMX4–IRES–rCD2 containing TLR9 wt or mutants were used for viral transduction of the mBa/F3 cell line and TLR9-knockout mouse bone marrow (BM) cells. Cells were treated with different TLR9 agonists: hODN10104, hODN2006, and minH75 (minH) (21) and mODN1826, mODN1668, and minM80 (minM) (20) (all B-class ODNs), all with a phosphorothioate (PTO) backbone (if not stated otherwise). The other minimal ODNs had a phosphodiester (PD) backbone (all from Sigma or Integrated DNA Technologies). The C-class ODN2395 and its variant C22 had a PD backbone. gDNA used for TLR9 activation was isolated from bacteria using a DNeasy Blood & Tissue kit T (QIAGEN), according to the manufacturer’s protocols. DNA from calf thymus (Bt-gDNA) was from Sigma. gDNA was mixed with DOTAP (Roche) at defined ratios before being added to the cells.

For transduction of Ba/F3 cells, Plat-E cells were transfected with wt TLR9, TLR9 mutants, or TLR9 chimeras cloned into pMX–IRES–rCD2 or pMX4–IRES–rCD2 vectors (retroviral pMXs–IRES system was from Cell Biolabs) using PEI MAX reagent (Polysciences). On the day of transduction, 5 × 104 Ba/F3 cells were mixed with virus supernatants from Plat-E cells, IL-3, and DOTAP. Three days after transduction, the cells were treated with ODNs or Pam3CSK4 (positive control).

For transduction of conventional DCs (cDCs), BM was obtained from the femurs and tibias of wt and TLR9−/− mice. After RBC lysis, cells (1 × 106 cells/ml) were seeded onto a 10-cm dish in DC media (RPMI 1640, 10% FBS, Pen/Strep, 50 μM 2-ME), supplemented with GM-CSF (10 ng/ml). Plat-E cells were transfected with wt TLR9 and TLR9 mutants cloned into the pMX–IRES–rCD2 or pMX4–IRES–rCD2 vector. On the day of transduction, a day after isolation, 3 × 105 cDCs were mixed with virus supernatants from Plat-E cells, GM-CSF (10 ng/ml), and DOTAP. Cells were transduced three times in 24-h intervals. On day 7, cells were treated with ODNs or Pam3CSK4 (positive control).

Unmethylated DNA was produced from NIH-3T3 gDNA by two-step PCR using random hexamer primers (ReadyMade Randomers; IDT). For one reaction (25 μl), we used 120 ng of gDNA, 20 pmol/μl primers, 12.5 μl of 2× KAPA HiFi HotStart ReadyMix (Kapa Biosystems), and 1 μl of DMSO. The PCR regimen was 95°C for 3 min; 98°C for 1 min, 11°C for 1 min, and 72°C for 90 s (35 times); and 72°C for 5 min. After 35 cycles, 25 μl of fresh reaction mix without DNA was added (resulting in 50 μl final volume), and 45 additional cycles were run. The PCR regimen was 95°C for 2 min; 98°C for 30 s, 11°C for 30 s, and 72°C for 90 s (45 times); and 72°C for 5 min. Controls were reaction without DNA and reaction without primers. DNA concentration in samples was measured with a Quant-IT dsDNA Assay Kit, high sensitivity (Invitrogen), according to the manufacturer’s instructions.

Ba/F3 NF-κB–GFP cells transduced with TLR9 mutants or chimeras were seeded onto a U-shaped 96-well plate and stimulated with TLR9 ligands (mODN1668, mODN1826, and hODN2006) and a TLR2 ligand (Pam3CSK4). After 20 h, the cells were stained for rCD2 expression (anti-rCD2–PE, clone QX-34; BioLegend), and the stained cells were subjected to flow cytometry (FACSCalibur; BD Biosciences). rCD2+ cells were analyzed for GFP expression. Acquired data were analyzed with FlowJo software (TreeStar). Each experiment was repeated independently at least three times.

wt or TLR9−/− BM-derived cDCs (BMcDCs) transduced with TLR9 mutants were seeded onto a U-shaped 96-well plate and stimulated with TLR9 ligands (mODN1668, mODN1826, and hODN2006) and a TLR2 ligand (Pam3CSK4). After 20 h, the cells were stained for expression of rCD2 (anti-rCD2–PE, clone QX-34; BioLegend), and CD69 (anti-mCD69-allophycocyanin, clone H1.2F3; eBioscience) and analyzed by flow cytometry (LSRFortessa; BD Biosciences). rCD2+ cells were analyzed for CD69 expression. Acquired data were analyzed with FlowJo software (TreeStar). Each experiment was repeated independently at least three times.

HEK293 cells were harvested from an actively growing culture and plated onto Costar White 96-well plates (Corning) at 2.2 × 104 cells per well (0.1 ml). After 24 h at 50% confluence, cells were transiently transfected with plasmids expressing wt TLR9 (20 ng of DNA per well) or mutants, UNC93B1 (5 ng of DNA per well), ELAM1-luciferase reporter plasmid (50 ng of DNA per well), and phRL-TK (5 ng of DNA per well) (unless stated otherwise), using Lipofectamine 2000, according to the manufacturer’s instructions (Invitrogen). The total amount of DNA for each transfection was kept constant by adding the appropriate amounts of pcDNA3 (Invitrogen) plasmid. The culture medium was replaced with fresh medium 24 h later, and cells were stimulated with TLR9 agonists (1 or 10 μM, unless stated otherwise) for 18 h.

To analyze whether the loss-of-function mutations maintained their ability to bind ligands, we tested the effect of coexpression of mutants (4 and 10 ng of DNA per well) and TLR9HA (2 ng of DNA per well) on NF-κB activation, upon stimulation with 3 μM hODN10104.

Cells were lysed in a passive lysis buffer (Promega) and analyzed for reporter gene activities using a dual-luciferase reporter assay. Relative luciferase activity (relative light units) was calculated by normalizing each sample’s firefly luciferase activity to the constitutive Renilla luciferase activity determined in the same sample. Relative luciferase activity (NF-κB activity) was calculated relative to wt hTLR9/hODN2006 or wt mTLR9/mODN1826 output after the subtraction of intrinsic activity (relative light units of cells treated with buffer). Each experiment was repeated independently at least two times and performed in at least three biological parallels (mean ± SD). An unpaired Student two-tailed t test was used for statistical comparisons.

Ramos-Blue cells or RAW-Blue cells were seeded at a density of 2 × 105 cells per well onto Costar 96-well plates (Corning). The cells were immediately stimulated with ODNs (3 μM final concentration). Supernatants were collected 20 h later, and NF-κB/AP-1 activation linked to SEAP was determined using QUANTI-Blue reagent, according to the manufacturer’s instructions (Invitrogen). Error bars represent the SD obtained from three biological replicates. The data are representative of at least two independent experiments.

HEK293T cells were seeded onto 12-well plates (Techno Plastic Products) at 2.2 × 105 cells per well. After 24 h at 50% confluence, cells were transiently transfected with 1470 ng of DNA per well of plasmid expressing wt hTLR9HA or mutants and 5 ng of UNC93B1 and 10 ng of plasmid constructively expressing mCerulean (for transfection control), using Lipofectamine 2000 transfection reagent. Forty-eight hours after transfection, the cells were lysed with RIPA buffer (50 mM Tris [pH 7.5], 150 mM NaCl, 1% [v/v] Triton X-100, 0.1% SDS, 0.5% deoxycholate) with cOmplete, Mini Protease Inhibitor (Roche Applied Science), sonicated, and centrifuged. Proteins from the supernatant were separated by SDS-PAGE and transferred to a Hybond-ECL nitrocellulose membrane (GE Healthcare). The membrane was washed and incubated in a blocking buffer (1× PBS, 0.1% Tween-20, 0.2% I-Block; Tropix). Blots were incubated with the primary Abs rabbit anti-HA (1:1000; H6908; Sigma), rabbit anti-GFP (1:1000; A11122; Invitrogen), or mouse anti–β-actin (1:1000; 3700; Cell Signaling Technology), followed by incubation with secondary goat anti-mouse IgG-HRP Ab (sc-2005; Santa Cruz Biotechnology) and goat anti-rabbit IgG-HRP Ab (ab6721; Abcam), both diluted 1:4000. Secondary Abs were detected with ECL Western Blotting Detection Reagent (GE Healthcare), according to the manufacturer’s protocol.

The AlphaScreen assay is a bead-based nonradioactive binding assay system used to detect biomolecular interactions. Binding of the biological partners brings the donor and acceptor beads into close proximity, and AlphaScreen acceptor beads emit a broad signal between 520 and 620 nm. The assays were performed in a final reaction volume of 20 μl of assay buffer, which contained 50 mM MOPS-NaOH (pH 6.5), 100 mM NaCl, 0.01% (w/v) Tween-20, and 1 g/l of BSA, in a ProxiPlate-384 (PerkinElmer). ODN2006 labeled with biotin–tetraethylene glycol at the 3′ end (hODN2006-b) (Sigma) was used. In the assay, 4 μl of cell lysate (prepared as described below) was incubated with hODN2006-b (up to 75 μM final concentration) for 1 h at 23°C. Anti-HA acceptor beads (20 μg/ml final concentration) were added and incubated for 1 h at 30°C. Then, streptavidin-coated donor beads (20 μg/ml final concentration) were added and incubated for 30 min at 30°C before the signals were measured with EnVision (PerkinElmer). The binding of hODN2006 to wt and mutant TLR9 after subtraction of intrinsic activity (no hODN2006-b) is shown in AlphaScreen units (all n = 3, mean ± SD; two independent experiments).

HEK293T cells seeded onto a 24-well plate (Techno Plastic Products), at 1.5 × 105 cells per well, were transiently transfected with a phRL-TK plasmid expressing Renilla luciferase as a transfection control (10 μg of DNA per well), UNC93B1 (5 ng of DNA per well), and plasmid expressing hTLR9HA wt or mutants, hTLR3HA, hTLR8HA, or pcDNA3 plasmid (control) (735 μg of DNA per well), using Lipofectamine transfection reagent (Invitrogen). Cells were lysed 48 h later in 0.2 ml of lysis buffer (50 mM Tris-HCl [pH 7.4], 1 mM EDTA, 150 mM NaCl, 0.5% Triton X-100, 20% glycerol, 1 mM Na3VO4, 25 mM NaF, 1 mM PMSF) containing a cOmplete Mini Protease Inhibitor (Roche) for 10 min on ice, and cell lysates were centrifuged (10,000 rpm, 10 min, 4°C). The transfection efficiency was determined with Renilla luciferase activity. In AlphaScreen assays, amounts of cell extracts that contained equal Renilla luciferase activity were used.

Mouse BM was obtained by flushing femurs and tibias of 8–10-wk-old wt C57BL/6J and TLR9−/− C57BL/6J mice (from Harlan, Udine, Italy and Ifrec, University of Osaka, Osaka, Japan, respectively) with RPMI 1640 and 10% FBS. After lysis of RBCs with 0.88% (w/v) NH4Cl (in water) for 10 min at 37°C, cells (1–3 × 106 cells per milliliter) were cultured at 37°C for a week in media containing RPMI 1640, 10% FBS, and Flt3 ligand (30 ng/ml) (for mBM-derived DCs cultured with Flt3 ligand [mBM-DCFlt3]) with penicillin (100 U/ml) and streptomycin (0.1 mg/ml). On day 3, an equal volume of fresh medium was added to the cells. On day 6, immature mBM-DCFlt3 were harvested. mBM-DCFlt3 were cultured in 96-well round-bottom plates and stimulated with ODNs. mTNF-α and mIL-6 secreted in the media were determined by ELISA (mTNF-α and mIL-6 Ready-Set-Go! ELISA; eBioscience).

wt or TLR9−/− BMcDCs transduced with TLR9 mutants were seeded onto a U-shaped 96-well plate and stimulated with TLR9 agonists (mODN1668, mODN1826, and hODN2006) and a TLR2 agonist (Pam3CSK4). mTNF-α, mIL-12 p40, and mIL-6 secreted in the media was determined 20 h later by ELISA (mTNF-α, mIL-12 p40, and mIL-6 Ready-Set-Go! ELISA; eBioscience).

HEK293T cells were seeded onto eight-well tissue culture chambers (Ibidi) at 2 × 105 cells per well. Twenty-four hours later, cells were cotransfected with plasmids expressing the TLR9HA mutant (140 ng of DNA per well) and a chimeric wt hTLR9–YFP (80 ng of DNA per well). At 48 h posttransfection, cells were fixed and permeabilized. TLR9HA was stained using rabbit anti-HA Abs (H6908, 1:50; Sigma) and secondary anti-rabbit Abs [F(ab)2 fragment of goat anti-rabbit IgG (H+L)] conjugated with Alexa Fluor 647 (A21246, 2 mg/ml; Invitrogen). Successive images were captured for wt hTLR9-YFP and TLR9HA mutant excited at 514 nm (fluorescence emission 525–570 nm) and 633 nm (fluorescence emission 650–700 nm), respectively. Images were acquired using a Leica TCS SP5 inverted laser-scanning microscope on a Leica DMI 6000 CS module, equipped with a HCX Plan Apochromat Lambda blue 63× oil-immersion objective (NA 1.4) (Leica Microsystems). Images were processed with LAS AF software (Leica Microsystems) and ImageJ software (National Institute of Mental Health, Bethesda, MD).

TLR9 and TLR8 amino acid sequences were aligned using ClustalW software (http://embnet.vital-it.ch/software/ClustalW.html). An hTLR9 ECD (aa 26–827) structural model was generated using I-TASSER software (http://zhanglab.ccmb.med.umich.edu/I-TASSER/) (23, 24). The structures of TLR9 (5) were used as templates by I-TASSER.

UCSF Chimera 1.6.2 software was used to generate structural figures, calculate surface hydrophobicity and electrostatic potential (distance from surface 1.4 Å, dielectric constant 2.0), and determine the distances among atoms (http://www.cgl.ucsf.edu/chimera) (25). Graphs were prepared with Origin 8.1 (http://www.originlab.com), and GraphPad Prism 5 (http://www.graphpad.com) was used for statistics. All data shown are the mean (± SD) of at least four biological replicates. A Student unpaired two-tailed t test or an ANOVA test followed by the Tukey post hoc test was used for statistical comparisons of data. For all tests, p values < 0.05 were considered statistically significant.

TLR9 activation efficiency of species-specific stimulatory ODNs was compared in mouse cells (RAW macrophages, BM-DCFlt3) and human cells (Ramos B-lymphocytes). In addition to widely used ODNs (hODN2006, mODN1862), minimal sequence motif ODNs [minH and minM (20, 21)] were used to stimulate hTLR9 and mTLR9. minM has only a single CpG motif positioned 6 nt from the 5′ end, whereas minH has a pair of CpG motifs separated by 7 nt, with the first TCG trinucleotide positioned at the 5′ end. mODNs (mODN1826, minM) (20) activated RAW macrophages and BM-DCFlt3. hODNs (hODN2006, minH) (21) effectively activated Ramos cells, and they also activated mouse RAW macrophages and mBM-DCFlt3 to some degree (Fig. 1A, 1B).

FIGURE 1.

N-terminal TLR9 ECD determines the sequence-specific recognition of ODNs. (A) Human B lymphocytes (Ramos-Blue) and RAW macrophages (RAW-Blue) were stimulated with PTO-based mODNs (mODN1826, minM) or hODNs (hODN2006, minH). The alkaline phosphatase activity under control of the NF-κB promoter was measured 20 h after cell stimulation. Data are representative of two independent experiments; each symbol represents the mean of four biological replicates ± SD. (B) mBM-DCFlt3 were stimulated with ODNPTO (0.2 μM). The secretion of mTNF-α and mIL-6 was determined 14 h after cell treatment. Data are representative of two independent experiments; each bar represents the mean of three biological replicates ± SD. (C) Schematic diagram for transduction of the Ba/F3 NF-κB–GFP pro B cell line. Plat-E cells were transfected with pMX–IRES–rCD2 plasmid carrying TLR9s. Three days later, the supernatants of Plat-E cells with retroviruses were collected. Ba/F3 NF-κB–GFP cells were transduced with pMX–IRES–rCD2 carrying TLR9s. (D) Three days after transduction, Ba/F3 cells expressing TLR9s were stimulated with ODNPTO or Pam3CSK4 as a control. Twenty-four hours later, the synthesis of GFP, whose expression is under control of the NF-κB promoter, was analyzed in cells positive for the cell surface marker rCD2. Geometric mean fluorescence intensity (gMFI) for each experiment versus ODN concentration was calculated (bottom panels). Data are representative of two independent experiments. (E) HEK293 cells transfected with wt TLR9s or mECD-hTLR9 were stimulated with ODNs (0.2, 0.5, 1 μM). (F) HEK293 cells transfected with wt TLR9s and the chimeric receptor m504-hTLR9 were stimulated with ODNPTO (5 μM). In (E) and (F), NF-κB activity was measured 24 h after stimulation. The activities were calculated after subtraction of the background activity for each ODN. Data are representative of three independent experiments; each bar represents the mean of four biological replicates ± SD. Localization (G) and expression (H) of the chimeric proteins are equal to those of wt TLR9. (G) Forty-eight hours after the transfection of HEK293 cells, localization of wt hTLR9GFP (cyan) and the HA-tagged chimeric TLR9 (yellow) was analyzed. Scale bar = 10 μM. (H) The expression of wt TLR9 and mutants was detected using anti-HA Abs; internal control GFP was detected with anti-GFP Abs.

FIGURE 1.

N-terminal TLR9 ECD determines the sequence-specific recognition of ODNs. (A) Human B lymphocytes (Ramos-Blue) and RAW macrophages (RAW-Blue) were stimulated with PTO-based mODNs (mODN1826, minM) or hODNs (hODN2006, minH). The alkaline phosphatase activity under control of the NF-κB promoter was measured 20 h after cell stimulation. Data are representative of two independent experiments; each symbol represents the mean of four biological replicates ± SD. (B) mBM-DCFlt3 were stimulated with ODNPTO (0.2 μM). The secretion of mTNF-α and mIL-6 was determined 14 h after cell treatment. Data are representative of two independent experiments; each bar represents the mean of three biological replicates ± SD. (C) Schematic diagram for transduction of the Ba/F3 NF-κB–GFP pro B cell line. Plat-E cells were transfected with pMX–IRES–rCD2 plasmid carrying TLR9s. Three days later, the supernatants of Plat-E cells with retroviruses were collected. Ba/F3 NF-κB–GFP cells were transduced with pMX–IRES–rCD2 carrying TLR9s. (D) Three days after transduction, Ba/F3 cells expressing TLR9s were stimulated with ODNPTO or Pam3CSK4 as a control. Twenty-four hours later, the synthesis of GFP, whose expression is under control of the NF-κB promoter, was analyzed in cells positive for the cell surface marker rCD2. Geometric mean fluorescence intensity (gMFI) for each experiment versus ODN concentration was calculated (bottom panels). Data are representative of two independent experiments. (E) HEK293 cells transfected with wt TLR9s or mECD-hTLR9 were stimulated with ODNs (0.2, 0.5, 1 μM). (F) HEK293 cells transfected with wt TLR9s and the chimeric receptor m504-hTLR9 were stimulated with ODNPTO (5 μM). In (E) and (F), NF-κB activity was measured 24 h after stimulation. The activities were calculated after subtraction of the background activity for each ODN. Data are representative of three independent experiments; each bar represents the mean of four biological replicates ± SD. Localization (G) and expression (H) of the chimeric proteins are equal to those of wt TLR9. (G) Forty-eight hours after the transfection of HEK293 cells, localization of wt hTLR9GFP (cyan) and the HA-tagged chimeric TLR9 (yellow) was analyzed. Scale bar = 10 μM. (H) The expression of wt TLR9 and mutants was detected using anti-HA Abs; internal control GFP was detected with anti-GFP Abs.

Close modal

To confirm that species- and sequence-specific activation are a consequence of the sequence of ODN and origin of TLR9 and not of the cell types, we used the mouse B cell line Ba/F3 and HEK293 cells, which do not express functional TLR9. Ba/F3 cells transduced with wt TLR9 (Fig. 1C, 1D) and positive for the cell surface marker rCD2 were analyzed for the expression of a reporter GFP after stimulation with ODNs or the triacylated lipopeptide Pam3CSK4, an agonist of TLR1/2 (Fig. 1D). As expected, cells expressing mTLR9, but not hTLR9, were activated by mODN1826 at the tested concentrations. hODN2006 activated hTLR9-expressing cells, as well as mTLR9-expressing cells, although less effectively. Similar to hTLR9-transduced Ba/F3 cells, hTLR9-transfected HEK293 cells were strongly activated by minH and minimally activated by minM (Fig. 1E). In contrast, mTLR9-expressing HEK293 cells were activated by all tested ODNs; however, a less efficient response was determined for hODNs (minH). Taken together, reduced efficacy and specificity of mTLR9 activation are characteristic for B-type hODNs with a PTO backbone. In contrast, BMcDCs and Ba/F3 cells expressing hTLR9 were not activated by mODNs at the tested concentrations, confirming sequence-specific requirements of hTLR9 for ODN.

To rule out that species-specific differences might be due to the TIR domain, we prepared a chimeric TLR9 with mouse ECD linked to the human trans-membrane and the TIR domain (mECD-hTLR9). Properties of the chimeric mECD-hTLR9, when stimulated with hODNs or mODNs, were similar to mTLR9 (Fig. 1D, 1E), which confirmed that the species-specific activation of TLR9 is due to ODN recognition by the ECD and not due to the TIR domain signaling cascade.

To identify the region within ECD defining the minimal sequence requirements for hTLR9, we prepared a chimeric receptor combining segments of human and mouse ECD sequences. The N-terminal segment of hECD up to aa 503 comprising a previously identified N-site, but not a C-site, was replaced with mECD (aa 1–504). The chimeric receptor m504-hTLR9, for which cellular localization and expression were similar to those of wt TLR9s, was activated with hODNs and mODNs (Fig. 1F–H). This confirms that activation of hTLR9 requires ODNs with sequence motifs that differ from those activating mTLR9, with regard to the number and position of CpG motifs, as well as that the N-terminal region of TLR9 ECD governs the sequence-specific activation of hTLR9.

Considering the binding sites determined by the crystal structure of mTLR9 with ligand, it is unlikely that these sites define the sequence specificity, because the amino acid residues that bind the CpG motif are conserved between mTLR9 and hTLR9 (5). In an effort to precisely identify the residues within the ECD that are responsible for the sequence-specific discrimination between hODNs and mODNs, we focused on the TLR9 region N-terminal to the Z-loop of TLR9 ECD. Based on the homology between TLR9 and TLR8, the residues selected for mutations were guided by the TLR8 corresponding to the binding sites of TLR8 (6, 7).

Two conserved sites (site A and site B) were identified that are strongly enriched in aromatic residues and surrounded by positively charged amino acid residues, which are characteristic of interactions with polynucleotides. To investigate the importance of site A, we selected the conserved residues Y345, S350, F375, F402, and R426 for point mutations (Fig. 2A). Mutations of aromatic residues Y345, F375, and F402 to alanine led to the almost complete loss of TLR9 activity (Fig. 2B). The S350A substitution had a minor effect, whereas the R426A mutation had no effect on TLR9 activity. The set of investigated residues was extended to include Y321 and Q346, as well as positively charged R348 and R377 surrounding the aromatic cluster. The Y321A mutation actually improved TLR9 activation upon stimulation with hODN10104, mutations Q346A and R348A decreased TLR9 activity by half, and R377A rendered TLR9 almost completely inactive (Fig. 2B).

FIGURE 2.

Residues within sites A and B that face each other in the dimer affect activation of TLR9. (A) A model of hTLR9 ECD is shown as a solvent-exposed surface and colored according to hydrophobicity. Amino acid residues within sites A (cyan) and B (magenta) selected for mutations are shown in detail. The residues whose mutations have a detrimental effect on TLR9 activity are highlighted in yellow, and the others are highlighted in green. Mutations in site A (B) and site B (C) residues affect TLR9 activity. HEK293 cells expressing wt hTLR9 or mutants were stimulated with hODN10104PTO (3 μM), and NF-κB–dependent luciferase activity was measured. The relative activities of mutants compared with wt TLR9 were calculated after subtraction of the background activity for each ODN. Data are representative of three independent experiments; each bar represents the mean of four biological replicates ± SD. Cyan and magenta lines indicate site A and site B amino acids, respectively. (D) hTLR9HA mutants bind hODN2006-3′ biotin (20, 40, 60 nM) similarly to wt hTLR9HA. The AlphaScreen method with a streptavidin donor and anti-HA acceptor beads was used to measure ODN binding to TLR9. Cell extracts obtained from HEK293 cells transfected with the pcDNA3 plasmid (control) and hTLR8HA were used as controls. Data are representative of two independent experiments. Each bar represents the mean of three technical replicates ± SD. (E) Inhibition of wt TLR9 (2 ng) activation by coexpression of TLR9 mutants. HEK293 cells cotransfected with plasmids expressing wt and mutant TLR9HA (4 and 10 ng) were stimulated with hODN10104PTO (3 μM), and NF-κB–dependent luciferase activity was measured. Data are representative of two independent experiments; each bar represents the mean of four biological replicates ± SD. Expression (F) and localization (G) of hTLR9 mutants compared with wt TLR9. (F) Western blot analysis of wt and mutant TLR9HA. wt TLR9 and mutants were detected using anti-HA Abs; actin as an internal control was detected with anti-actin Abs. (G) Microscopic images of hTLR9HA mutants (cyan) and wt hTLR9-GFP (yellow) expressed in HEK293 cells. Scale bar = 10 μM. **p < 0.05. ns, p > 0.05.

FIGURE 2.

Residues within sites A and B that face each other in the dimer affect activation of TLR9. (A) A model of hTLR9 ECD is shown as a solvent-exposed surface and colored according to hydrophobicity. Amino acid residues within sites A (cyan) and B (magenta) selected for mutations are shown in detail. The residues whose mutations have a detrimental effect on TLR9 activity are highlighted in yellow, and the others are highlighted in green. Mutations in site A (B) and site B (C) residues affect TLR9 activity. HEK293 cells expressing wt hTLR9 or mutants were stimulated with hODN10104PTO (3 μM), and NF-κB–dependent luciferase activity was measured. The relative activities of mutants compared with wt TLR9 were calculated after subtraction of the background activity for each ODN. Data are representative of three independent experiments; each bar represents the mean of four biological replicates ± SD. Cyan and magenta lines indicate site A and site B amino acids, respectively. (D) hTLR9HA mutants bind hODN2006-3′ biotin (20, 40, 60 nM) similarly to wt hTLR9HA. The AlphaScreen method with a streptavidin donor and anti-HA acceptor beads was used to measure ODN binding to TLR9. Cell extracts obtained from HEK293 cells transfected with the pcDNA3 plasmid (control) and hTLR8HA were used as controls. Data are representative of two independent experiments. Each bar represents the mean of three technical replicates ± SD. (E) Inhibition of wt TLR9 (2 ng) activation by coexpression of TLR9 mutants. HEK293 cells cotransfected with plasmids expressing wt and mutant TLR9HA (4 and 10 ng) were stimulated with hODN10104PTO (3 μM), and NF-κB–dependent luciferase activity was measured. Data are representative of two independent experiments; each bar represents the mean of four biological replicates ± SD. Expression (F) and localization (G) of hTLR9 mutants compared with wt TLR9. (F) Western blot analysis of wt and mutant TLR9HA. wt TLR9 and mutants were detected using anti-HA Abs; actin as an internal control was detected with anti-actin Abs. (G) Microscopic images of hTLR9HA mutants (cyan) and wt hTLR9-GFP (yellow) expressed in HEK293 cells. Scale bar = 10 μM. **p < 0.05. ns, p > 0.05.

Close modal

Within the TLR9 dimer, site A faces site B of the opposite TLR9, enabling sites A and B to form a hydrophobic pocket in a TLR9 dimer. Within proposed site B, we selected for investigation the residues D534 and Y536 (as control), which correspond to the D535 and Y537 residues of mTLR9 that were previously proposed to participate in binding CpG (26), as well as residues K532, F559, G560, and Q562 (Fig. 2A), all of which are positioned in the vicinity of the ligand-binding site of TLR8 (6). Substitutions D534A, Y536A, K532D, and F599A rendered TLR9 inactive (Fig. 2C), whereas mutations G560A and Q562A reduced TLR9 activation by hODN10104. To evaluate whether ODN binding is affected by the mutations of hTLR9, we used an AlphaScreen assay to determine binding of hODN2006 (3′-biotin labeled) to the HA-tagged wt hTLR9 and mutants (Fig. 2D). ODN binding was not appreciably impaired by single-point mutations, demonstrating that ODN binding alone is not sufficient for TLR9 activation, because the mutations might affect the formation of active dimers. Although the residues selected for mutagenesis were located at the surface of the receptor in the homology model of hTLR9, we investigated the effect of mutants on wt TLR9 NF-κB activation to rule out inactivation of mutants due to the loss of their structural integrity as a result of the misfolding, as well as to analyze whether the loss-of-function mutations affect their ability to bind ODN (Fig. 2E). Inactive TLR9 mutants that retain ODN-binding ability should inhibit activation of wt TLR9 due to the sequestration of ODN, whereas folding-defective mutants should have no effect. Indeed, all tested signaling-incompetent TLR9 mutants at site B, with the exception of G560A, were also able to inhibit wt TLR9 signaling. Expression levels and cellular localization of TLR9 mutants were comparable to those of wt TLR9 (Fig. 2F, 2G). These results indicate that amino acid residues within proposed sites A and B participate in TLR9 activation.

After identifying the region of hTLR9 ECD important for sequence-specific recognition of ODNs, we aimed to pinpoint the specific amino acid residues that are responsible for this effect. Residues Gln346 and Arg348, located near proposed site A, N-terminal to the Z-loop, that differ between human and mouse, are surface exposed and face site B in the dimer (Supplemental Fig. 1), although they are too far apart to interact directly with another ECD. These two residues were substituted with the mTLR9 counterpart residues, replacing polar Gln346 with a positively charged arginine and the positively charged Arg348 with a positively charged lysine. Because the selected amino acid residues are positioned close to one another, a synergistic effect of these two amino acid residues might occur; therefore, a double mutant Q346R/R348K was prepared, substituting both residues with mTLR9 counterparts. Experiments in a Ba/F3 cell line expressing a GFP reporter under control of the NF-κB promoter and transduced with the hTLR9 Q346R/R348K mutant demonstrated that hTLR9 with a double hQ346R/R348K mutation gained responsiveness to mODNs (Fig. 3A). Results were confirmed using BMcDCs from TLR9−/− mice. BMcDCs were transduced with retroviruses carrying wt TLR9s or a double hQ346R/R348K mutant (Fig. 3B depicts the experimental scheme of BMcDC transduction). Transduced BMcDCs were stimulated with mODNs or hODNs, and expression of TNF-α, IL-6, and IL-12p40 was analyzed (Fig. 3C, 3D, Supplemental Fig. 2A, a dose response of BMcDCs expressing wt TLR9s and hQ346R/R348K mutant). hQ346R/R348K hTLR9+ BMcDCs (rCD2+) released cytokines and expressed CD69 at the cell surface after stimulation with mODNs, which was not the case for wt hTLR9 BMcDCs. These results suggest that residues Q346 and R348 restrict the hTLR9-specific response to the pattern that is characteristic of ODNs able to stimulate hTLR9.

FIGURE 3.

Residues Q346 and R348 participate in defining the ODN specificity of hTLR9. (A) Ba/F3 NF-κB–GFP cells transduced with wt hTLR9, hQ346R/R348K, or an empty vector as a negative control were stimulated with hODNPTO or mODNPTO or a TLR2 agonist Pam3CSK4 as a positive control. GFP expression was analyzed in rCD2+ cells. The efficiency of transduction was established via analysis of cells positive for rCD2. Twenty-four hours after the stimulation of cells with ODNPTO or Pam3CSK4, the expression of GFP in rCD2+ cells was analyzed. Data are representative of two independent experiments (see also Supplemental Fig. 2A). (B) Schematic diagram of transduction of BMcDCs from BM of TLR9−/− mice. BMcDCs were transduced on days 1, 2, and 3 with a pMX vector carrying TLR9s. On day 7, BMcDCs were stimulated with ODNs or Pam3CSK4. The TLR9 response in rCD2+ BMcDCs was analyzed 24 h after stimulation. (C) BMcDCs transduced with wt TLR9s and the hQ346R/R348K double mutant were stimulated with ODNPTO or Pam3CSK4. Twenty-four hours after stimulation, the TLR9 response was analyzed by the expression of cell surface marker CD69 in rCD2+ BMcDCs. Geometric mean of GFP fluorescence intensity (gMFI) was determined for the rCD2+ population. Data are representative of two independent experiments (see also Supplemental Fig. 2A). (D) Twenty-four hours after stimulation with ODNPTO or Pam3CSK4, the expression of mTNF-α, mIL-6, and mIL-12p40 was determined in BMcDC transduced with wt TLR9s and the hQ346R/R348K mutant. Transduction efficiency is indicated as the percentage of rCD2+ cells determined by flow cytometry. Data are representative of two independent experiments. Each bar represents the mean of three biological replicates ± SD.

FIGURE 3.

Residues Q346 and R348 participate in defining the ODN specificity of hTLR9. (A) Ba/F3 NF-κB–GFP cells transduced with wt hTLR9, hQ346R/R348K, or an empty vector as a negative control were stimulated with hODNPTO or mODNPTO or a TLR2 agonist Pam3CSK4 as a positive control. GFP expression was analyzed in rCD2+ cells. The efficiency of transduction was established via analysis of cells positive for rCD2. Twenty-four hours after the stimulation of cells with ODNPTO or Pam3CSK4, the expression of GFP in rCD2+ cells was analyzed. Data are representative of two independent experiments (see also Supplemental Fig. 2A). (B) Schematic diagram of transduction of BMcDCs from BM of TLR9−/− mice. BMcDCs were transduced on days 1, 2, and 3 with a pMX vector carrying TLR9s. On day 7, BMcDCs were stimulated with ODNs or Pam3CSK4. The TLR9 response in rCD2+ BMcDCs was analyzed 24 h after stimulation. (C) BMcDCs transduced with wt TLR9s and the hQ346R/R348K double mutant were stimulated with ODNPTO or Pam3CSK4. Twenty-four hours after stimulation, the TLR9 response was analyzed by the expression of cell surface marker CD69 in rCD2+ BMcDCs. Geometric mean of GFP fluorescence intensity (gMFI) was determined for the rCD2+ population. Data are representative of two independent experiments (see also Supplemental Fig. 2A). (D) Twenty-four hours after stimulation with ODNPTO or Pam3CSK4, the expression of mTNF-α, mIL-6, and mIL-12p40 was determined in BMcDC transduced with wt TLR9s and the hQ346R/R348K mutant. Transduction efficiency is indicated as the percentage of rCD2+ cells determined by flow cytometry. Data are representative of two independent experiments. Each bar represents the mean of three biological replicates ± SD.

Close modal

Next, we evaluated the contribution of individual amino acid residues in defining the sequence specificity of hTLR9. For that task, we used HEK293 cells transfected with wt hTLR9 and mutants hQ346R, hR348K, and hQ346R/R348K. The hR348K mutation had a minor impact on activation with mODNs or hODNs, whereas the hQ346R and hQ346R/R348K mutants improved the response to mODNs compared with wt hTLR9. hQ346R/R348K also increased activation by hODNs (Supplemental Fig. 2B, 2C), while maintaining the expression and localization of mutants similar to wt TLR9 (Supplemental Fig. 2D, 2E). Taken together, Q346 was the main contributor to ODN specificity, with an additional contribution by R348.

For further analysis, we examined the possibility that residues facing hQ346R and hR348K (site A) within the TLR9 dimer might also affect sequence-specific responses. Selection of amino acid residues for point mutations was also based on the sequence alignment of TLR9 ECDs and on the molecular model of hTLR9 dimer based on the crystal structure of mTLR9 with agonist (5) (Supplemental Fig. 1). Residues in leucine-rich repeats (LRRs) 16, 17, and 18 were selected that differ between mTLR9 and hTLR9 and that face hQ346 and hR348. Amino acid residues (hG560, hQ562, and hV564) on LRR18 were mutated between human and mouse counterparts. In an initial screen, the effect of mutations was analyzed on HEK293 cells expressing wt TLRs and mutants (Fig. 4A, 4B). The hQ562K mutation slightly improved TLR9 activation by hODNs (Fig. 4A). The other two substitutions and a triple mutation from hTLR9 to mTLR9 residues had a minor impact on TLR9 activity (Fig. 4B). Expression levels and localization of TLR9 mutants were comparable to those for wt TLR9 (Supplemental Fig. 2D, 2E). The impact of hQ562K substitution on Ba/F3 cells suggests that residue hQ562 plays a minor role in the recognition of ODNs by TLR9, probably by increasing the efficacy of the receptor (Fig. 4C, Supplemental Fig. 2F, a dose response of Ba/F3 cells expressing hQ562K mutant).

FIGURE 4.

Residue Q562 contributes to hTLR9 ODN selectivity. (A and B) Characterization of mutations within proposed site B on TLR9 activation. HEK293 cells transfected with wt TLR9, as well as the mutants, were stimulated with ODNPTO, and NF-κB–driven luciferase activity was measured 24 h later. NF-κB luciferase (A) or relative NF-κB luciferase (B) activity after subtraction of background activity. Data are representative of two (A) or three (B) independent experiments. Each point represents the mean of three biological replicates ± SD, and the bars represents the mean of four biological replicates. (C) Ba/F3 NF-κB–GFP cells transduced with wt TLR9, hQ562K, and an empty vector as a negative control were stimulated with ODNPTO, as well as Pam3CSK4 as a positive control. GFP expression was analyzed in rCD2+ cells. The efficiency of transduction was established by analysis of rCD2+ cells. Expression of GFP in rCD2+ cells was analyzed 24 h after the stimulation of cells with ODNs or Pam3CSK4. Geometric mean of GFP fluorescence intensity (gMFI) was determined on the rCD2+ population (gMFI of nonstimulated cells is shown in parentheses). Data are representative of three independent experiments (see also Supplemental Fig. 2F). *p < 0.1.

FIGURE 4.

Residue Q562 contributes to hTLR9 ODN selectivity. (A and B) Characterization of mutations within proposed site B on TLR9 activation. HEK293 cells transfected with wt TLR9, as well as the mutants, were stimulated with ODNPTO, and NF-κB–driven luciferase activity was measured 24 h later. NF-κB luciferase (A) or relative NF-κB luciferase (B) activity after subtraction of background activity. Data are representative of two (A) or three (B) independent experiments. Each point represents the mean of three biological replicates ± SD, and the bars represents the mean of four biological replicates. (C) Ba/F3 NF-κB–GFP cells transduced with wt TLR9, hQ562K, and an empty vector as a negative control were stimulated with ODNPTO, as well as Pam3CSK4 as a positive control. GFP expression was analyzed in rCD2+ cells. The efficiency of transduction was established by analysis of rCD2+ cells. Expression of GFP in rCD2+ cells was analyzed 24 h after the stimulation of cells with ODNs or Pam3CSK4. Geometric mean of GFP fluorescence intensity (gMFI) was determined on the rCD2+ population (gMFI of nonstimulated cells is shown in parentheses). Data are representative of three independent experiments (see also Supplemental Fig. 2F). *p < 0.1.

Close modal

To evaluate whether the sequence specificity for ODNs of hTLR9 could be reversed from mTLR9 to hTLR9, mR346 and K348 were mutated to their human counterpart residues glutamine and arginine, respectively, and a double mutant R346Q/K348R-mTLR9 (mR346Q/K348R) was prepared. The effect of the mutations was analyzed on Ba/F3 cells and HEK293 cells expressing wt TLR9 or mTLR9 mutants. Although the replacement of R348 in hTLR9 had no effect, the reverse mutation in mTLR9 (mK348R) improved the response to hODNs compared with wt mTLR9 on Ba/F3 cells (Fig. 5A, Supplemental Fig. 3A, a dose response of Ba/F3 cells expressing mK348R). In contrast, mR346Q and the double mutation mR346Q/K348R had no significant impact on stimulation by hODNs. The observed effects were confirmed in HEK293 cells expressing mTLR9 variants with hODN2006 and minH ODN (Supplemental Fig. 3B–D), as well as with additional cytokine readout using BMcDCs transduced with wt TLR9s or mutants (Fig. 5B). mK348R enhanced the synthesis of TNF-α, IL-6, and IL-12p40 in response to hODNs and mODNs. The results show the context-dependent effect of mutations as a result of the participation of several sites in the formation of the signaling complex.

FIGURE 5.

Mutation K563Q of mTLR9 improved the activation with hODN. (A) Ba/F3 NF-κB–GFP cells (TLR9−/−) were transduced with wt TLR9s and mR346Q, mK348R, mR346Q/K348R, and mK563Q mutants. The transduced cells were stimulated with ODNs. The expression of NF-κB–driven GFP cells positive for the cell surface marker rCD2 was analyzed. Pam3CSK4 was used as a positive control. Geometric mean of fluorescence intensity (gMFI) was determined for the rCD2+ population. Data are representative of three independent experiments (see also Supplemental Fig. 3A). (B) BMcDCs isolated from TLR9−/− mice were transduced with wt TLR9s and mutants. Expression of TNF-α, IL-6, and IL-12p40 was determined 24 h after stimulation with ODNs (25, 50 nM). Transfection efficiency is indicated as the percentage of rCD2+ cells, as determined by flow cytometry. Data are representative of two independent experiments; each bar represents the mean of three biological replicates ± SD. gMFI of unstimulated cells is denoted in brackets.

FIGURE 5.

Mutation K563Q of mTLR9 improved the activation with hODN. (A) Ba/F3 NF-κB–GFP cells (TLR9−/−) were transduced with wt TLR9s and mR346Q, mK348R, mR346Q/K348R, and mK563Q mutants. The transduced cells were stimulated with ODNs. The expression of NF-κB–driven GFP cells positive for the cell surface marker rCD2 was analyzed. Pam3CSK4 was used as a positive control. Geometric mean of fluorescence intensity (gMFI) was determined for the rCD2+ population. Data are representative of three independent experiments (see also Supplemental Fig. 3A). (B) BMcDCs isolated from TLR9−/− mice were transduced with wt TLR9s and mutants. Expression of TNF-α, IL-6, and IL-12p40 was determined 24 h after stimulation with ODNs (25, 50 nM). Transfection efficiency is indicated as the percentage of rCD2+ cells, as determined by flow cytometry. Data are representative of two independent experiments; each bar represents the mean of three biological replicates ± SD. gMFI of unstimulated cells is denoted in brackets.

Close modal

Furthermore, we also examined the possibility that residues within site B of mTLR9 also affect sequence-specific responses. Amino acid residues (mS561, mK563, and mI565) on LRR18 were mutated from mouse to human counterparts. The effect of mutations was analyzed on HEK293 cells expressing wt TLRs and mutants (Supplemental Fig. 3C–E). Only the mK563Q mutation improved the TLR9 response to hODNs. mK563Q on Ba/F3 cells and on BMcDCs exhibited increased responsiveness to mODNs and gained responsiveness to hODNs (Fig. 5, Supplemental Fig. 3A, a dose response of Ba/F3 cells expressing mK563Q). These results suggest that the residue mK563 plays a role in receptor activation and that its role is in limiting the receptor activity rather than causing species selectivity.

We showed that the amino acid residues Q346 and R348 located N-terminal to the Z-loop within site A determine the sequence specificity for hTLR9 and that Q562 changes the sensitivity (Supplemental Table I summarizes the effects of mutations on TLR9 activity). hTLR9 with all three substituted amino acid residues responded to the PTO-based B-class mODNs as efficiently as did mTLR9 (Fig. 6A). The nuclease-resistant PTO backbone hinders degradation by nucleases. In contrast, PTO modifications lead to increased nonspecific binding to proteins (27, 28). We showed previously (21) that the sequence specificity of hTLR9 is preserved, regardless of the chemistry of the ODN backbone; this is not the case for mTLR9, whose sequence specificity is less stringent for PD-based ODNs (20). Therefore, we also analyzed the effect of Q346, R348, and Q562 substitutions on hTLR9 activation by PD-based mODNs: minM and mODN1826. Similar to PTO-based B-class mODNs, triple substitution sensitized hTLR9 to PD-based minM (Fig. 6A), which was also confirmed for PD-based mODN1826 (Fig. 6B). Taken together, the selected mutations improved efficacy and redefined the specificity of hTLR9 for mODNs, but with a lower fold change for PD-based ODNs compared with PTO-based ODNs. Mutations weakly improved efficacy of the receptor for hODNs (Fig. 6C), which is probably a consequence of the hQ562K substitution (Fig. 4).

FIGURE 6.

The mutations render hTLR9 more sensitive to C-class ODNs. HEK293 cells transfected with wt TLR9s or mutants were stimulated with mODNs (minM, mODN1826) (A and B) or hODNs (minH, hODN2006) (C) with a PTO or PD backbone. The NF-κB–driven luciferase activity was measured 24 h later. Data are representative of two independent experiments; each point represents the mean of three biological replicates ± SD. (D and E) HEK293 cells transfected with wt TLR9s were stimulated with C-class ODN2395PD and its variant C22PD, which specifically activate mTLR9 but not hTLR9. The NF-κB–driven luciferase activity was measured. Data are representative of three (D) or two (E) independent experiments; the mean of four biological replicates ± SD (D) or three biological replicates ± SD (E) are shown. **p < 0.05.

FIGURE 6.

The mutations render hTLR9 more sensitive to C-class ODNs. HEK293 cells transfected with wt TLR9s or mutants were stimulated with mODNs (minM, mODN1826) (A and B) or hODNs (minH, hODN2006) (C) with a PTO or PD backbone. The NF-κB–driven luciferase activity was measured 24 h later. Data are representative of two independent experiments; each point represents the mean of three biological replicates ± SD. (D and E) HEK293 cells transfected with wt TLR9s were stimulated with C-class ODN2395PD and its variant C22PD, which specifically activate mTLR9 but not hTLR9. The NF-κB–driven luciferase activity was measured. Data are representative of three (D) or two (E) independent experiments; the mean of four biological replicates ± SD (D) or three biological replicates ± SD (E) are shown. **p < 0.05.

Close modal

We also examined whether activation of hTLR9 with C-class ODNs is improved when Q346R, R348K, and Q562K mutations are introduced. The sequences of C-class ODNs (29) combine CpG motifs of B-class ODNs and the 3′ palindromic CG region, which forms intermolecular dimers under physiological conditions. The C-class ODN ODN2395 activates mTLR9 and hTLR9 but with more efficient activation of mTLR9 (Fig. 6D). We prepared a C-class ODN variant of ODN2395, in which the 5′ CG dinucleotide was substituted with CC (C22). These substitutions change the mixed species specificity of ODN2395 to ODNs that specifically activate mTLR9 but not hTLR9 (Fig. 6D). The Q346R mutation renders hTLR9 susceptible to the C22 variant (Fig. 6E). Additional substitutions further increase the sensitivity to C-class ODNs. The activation of mutated hTLR9 stimulated with C-class ODNs was not as high as that for mTLR9; nonetheless, hTLR9 with the Q346R, R348K, and Q562K mutations gained sensitivity for C-class mODNs. Taken together, these results support the idea that Q346 with some input from R348 define the two-CpG motif requirement, and Q562 determines sensing sensitivity.

In general, activation of hTLR9 requires a pair of CpG motifs separated by 6–10 nt. The 5′ CpG motif binds to the binding sites determined by the crystal structure of TLR9 with 12-nt ODN1668 (5). The second CpG motif facilitates activation of hTLR9 (Fig. 7A). In contrast, mODNs with a single CpG motif fail to form the active hTLR9–mODN complex (Fig. 7B). The substitution of Q346 and R348 in mouse counterparts enables activation of hTLR9 also by mODNs, whose sensitivity is potentiated by the additional substitution of Q562 (Fig. 7C) because it is characteristic of the mTLR9–mODN complex (Fig. 7D).

FIGURE 7.

Substitutions of Q346, R348, and Q562 in ECD of hTLR9 with mouse counterparts alter the sequence specificity of hTLR9. (A) hTLR9 and hODNs with two spaced CpG motifs form an active TLR9–ODN complex. (B) hTLR9 and mODNs with a single CpG motif form an inactive TLR9–ODN complex. (C) Substitutions hQ346R, R348K, and Q562K of hTLR9 and mODN form an active TLR9–ODN complex. (D) mTLR9 mODNs form an active TLR9–ODN complex.

FIGURE 7.

Substitutions of Q346, R348, and Q562 in ECD of hTLR9 with mouse counterparts alter the sequence specificity of hTLR9. (A) hTLR9 and hODNs with two spaced CpG motifs form an active TLR9–ODN complex. (B) hTLR9 and mODNs with a single CpG motif form an inactive TLR9–ODN complex. (C) Substitutions hQ346R, R348K, and Q562K of hTLR9 and mODN form an active TLR9–ODN complex. (D) mTLR9 mODNs form an active TLR9–ODN complex.

Close modal

We determined that PD- or PTO-based ODNs with at least two closely separated CpG motifs are efficient hTLR9 agonists, whereas ODN with a single CpG motif effectively activates mTLR9 (Figs. 1, 8A). The sequence specificity of TLR9 for PTO-based agonists differs between mice and humans, as well as among primates (17). Moreover, the minimal sequence-specific requirements for hTLR9 are independent of backbone chemistry; this is not the case for mTLR9, for which the sequence specificity for PTO-based ODNs is lost when using PD-based ODNs (20, 21). The minimal distance between the CpG motifs required for strong activation is ≥4 nt; the potency decreases if the distance is >10 nt, to only 30% at a separation of 28 nt (Fig. 8A). CpG is strongly underrepresented in the human genome in comparison with the bacterial genome (30). Moreover, a large fraction of hCpG motifs are methylated, which, in addition to the G-rich inhibitory sequences present in mammalian gDNA, was proposed as the basis for TLR9 selectivity for microbial DNA (9, 12). The requirement for a pair of CpGs separated by a defined range of nucleotides in hTLR9 further amplifies the selectivity bias for the differentiation between microbial and mammalian DNA. Analysis of the frequency of pairs of CpGs separated by 4–28 nt in several genomes demonstrated that the frequency increased in a nonlinear fashion, with >30 double CpG motifs per kilobase for Mycobacterium tuberculosis (66% CGs [CG content] and 13% CpGs [130 CpGs per kilobase]), decreasing to 6 double CpG motifs per kilobase for the genome of Escherichia coli (51% CG and 7.3% CpGs). In both cases, the frequency of CpG pairs in the genome was higher than in the randomly permutated nucleotide sequence. We also observed an interesting periodicity pattern of the motif count as a function of CpG separation in bacterial genomes, probably due to the neighboring codon bias (Fig. 8B). In contrast, hDNA exhibited <0.4 CpG pair motifs per kilobase, which was significantly lower than in the permutated sequence and demonstrated the bias against CpG in the human genome (Fig. 8C). This results in a selectivity ratio of up to 1:100, to which an additional factor of 25 can be added because of 70% of CpGs are methylated in mammalian genomes (31). Altogether, the sequence-specific requirements of hTLR9, the methylation of CpG motifs, and the G-rich inhibitory sequences (9) define activation efficiency of mammalian gDNA.

FIGURE 8.

hTLR9 recognizes the characteristically spaced pair of CpGs that determines the differentiation in sensing bacterial and mammalian DNA. (A) Effect of the separation between CpG motifs of ODNs on hTLR9 activation. RAMOS-Blue cells were stimulated with ODNPD variants having a 4–24-nt separation of the CpG motifs. Relative NF-κB activities were calculated after subtraction of the background activity for each ODN. (B and C) The frequency of sequences comprising two closely separated CpG motifs per kilobase in genomes from different organisms, depending on the distance between the CpG motifs. For hDNA, 70% CG methylation frequency was used for calculations. Data show the numbers of defined motifs detected within the genome sequence (B) and the values for a randomly permutated genome sequence (C). Human ch. 20 (human, CG frequency 0.44); Helicobacter pylori (CG frEquation 0.39); E. coli (CG frEquation 0.51); and M. tuberculosis (CG frEquation 0.66). (D) gDNA (7.5 ng/μl) isolated from bacteria stimulates mTLR9 better than hTLR9. Mutations hQ346R, hR348K, and hQ562K alter hTLR9 sensitivity to Bt-gDNA (E), unmethylated NIH3T3 gDNA (F), and bacterial gDNA (7.5 ng/μl) (G). In (F), fold changes are shown as the ratio of DNA-treated/PBS-treated HEK293 cells. HEK293 cells expressing wt TLR9s or mutants were stimulated with a suspension of gDNA and DOTAP (ratio 1:3). NF-κB–driven reporter activity was determined after 20 h. Data are representative of two independent experiments; each bar represents the mean of four biological replicates ± SD. **p < 0.05.

FIGURE 8.

hTLR9 recognizes the characteristically spaced pair of CpGs that determines the differentiation in sensing bacterial and mammalian DNA. (A) Effect of the separation between CpG motifs of ODNs on hTLR9 activation. RAMOS-Blue cells were stimulated with ODNPD variants having a 4–24-nt separation of the CpG motifs. Relative NF-κB activities were calculated after subtraction of the background activity for each ODN. (B and C) The frequency of sequences comprising two closely separated CpG motifs per kilobase in genomes from different organisms, depending on the distance between the CpG motifs. For hDNA, 70% CG methylation frequency was used for calculations. Data show the numbers of defined motifs detected within the genome sequence (B) and the values for a randomly permutated genome sequence (C). Human ch. 20 (human, CG frequency 0.44); Helicobacter pylori (CG frEquation 0.39); E. coli (CG frEquation 0.51); and M. tuberculosis (CG frEquation 0.66). (D) gDNA (7.5 ng/μl) isolated from bacteria stimulates mTLR9 better than hTLR9. Mutations hQ346R, hR348K, and hQ562K alter hTLR9 sensitivity to Bt-gDNA (E), unmethylated NIH3T3 gDNA (F), and bacterial gDNA (7.5 ng/μl) (G). In (F), fold changes are shown as the ratio of DNA-treated/PBS-treated HEK293 cells. HEK293 cells expressing wt TLR9s or mutants were stimulated with a suspension of gDNA and DOTAP (ratio 1:3). NF-κB–driven reporter activity was determined after 20 h. Data are representative of two independent experiments; each bar represents the mean of four biological replicates ± SD. **p < 0.05.

Close modal

To experimentally evaluate the differences in the stimulatory capacity of bacterial gDNA on mTLR9 and hTLR9, we used DNA isolated from Staphylococcus aureus (2.5% CpG and 0.8% TCG trinucleotide), E. coli (7.3% CpG and 1.5% TCG), and Pseudomonas aeruginosa (12% CpG and 2.71% TCG) (32), as well as Bt-gDNA and unmethylated DNA obtained by the PCR amplification of gDNA from NIH-3T3 cells. Activation of hTLR9 by gDNA was augmented by the introduction of DNA into cells by DOTAP, because it improves internalization of gDNA and activation of DCs and macrophages (32, 33). HEK293 cells expressing wt TLR9s and mutants were used. As a control, HEK293 cells expressing no TLR9 were used to discount activation of the cytosolic DNA-sensing system, such as cGAS-STING and AIM2 (34, 35). DNA from E. coli and P. aeruginosa elicited strong TLR9 activation (Fig. 8D), followed by the DNA from S. aureus, consistent with its lower CpG content. Mammalian gDNA (Bt-gDNA) only weakly activated hTLR9 and mTLR9, even in the presence of DOTAP (Fig. 8E). hTLR9 mutations hQ346R, hQ346R/hR348K, and hQ346R/hR348K/Q562K demonstrated augmented activation by Bt-gDNA in comparison with wt hTLR9 (Fig. 8E). Furthermore, the impact of hQ346, R348, and Q562 on activation of TLR9 by unmethylated versus methylated gDNA was analyzed. Similar to Stacey et al. (9), we showed that unmethylated PCR-amplified mammalian gDNA was a better agonist of hTLR9 than was methylated gDNA (Fig. 8F). Similarly increased activation was observed for hTLR9 mutants stimulated with bacterial gDNA (Fig. 8G).

Taken together, these results demonstrate that the sequence specificity of hTLR9 contributes to the decreased activation of hTLR9 by gDNA, particularly the unmethylated form. This phenomenon depends on the frequency of the CpG motifs within the genomes of the pathogens and the host.

Recognition of DNA by TLR9 is connected with a potential risk to human health due to the presence of endogenous DNA, which could trigger unwanted autoimmune responses, especially associated with autologous IgGs (36). The probability of activation of TLR9 by endogenous DNA-degradation fragments is further decreased by methylation of CpG motifs. This study investigated the underlying mechanism of the recognition of the more stringent double CpG motif by hTLR9 compared with mTLR9.

Although the structure of hTLR9 has not been determined, the crystal structures of horse, mouse, and bovine TLR9 with agonists (5) revealed the binding mechanism and pattern of ODN recognition demonstrated by the 12-nt-long ssDNA. Agonists shorter than 19 nt are very poor activators of TLR9 (20, 21). Therefore, we propose that, together with the binding sites identified in the crystal structure, interactions with longer, more potent ODNs underlie efficient TLR9 activation. Significant similarity allows us to build a rather reliable model of the tertiary structure of hTLR9 based on the crystal structures of orthologs with known three-dimensional structures (5), with the exception of the Z-loop, and the loops at LRR11 and LRR18, which are within proposed sites A and B. Sites A and B may affect the interaction between ECDs in a TLR9 dimer, or they may interact with longer agonists, such as the second CpG motif of hODN. It is possible that the residues identified in this study define the strength of the interaction between opposing TLR9s in a dimer and indirectly affect the sequence-specific requirements of hTLR9. We showed that Q346 and R348 N-terminal to the Z-loop, within site A, govern the agonist sequence specificity for hTLR9, with Q346 being the major contributor. Comparing the results and sequence alignment of site A, it seems that the stringent sequence specificity of TLR9 for type B ODNs is limited to humans and other primates (17). Glutamine at site 346 is conserved in TLR9 of humans and other primates, whereas mouse has arginine, and histidine is most common for other species. Mutations of the selected mouse amino acid residues to human counterparts had a minor impact on sequence specificity, which indicates that replacement of several amino acids may be needed for the two-CpG motif that is characteristic of hTLR9. Nevertheless, mutations mK348R and mK563Q augmented the activation of mTLR9 by all tested ODNs.

Amino acid residues hQ364, hR348, and hQ562 were identified as important for defining the response of hTLR9 to PTO- or PD-based mODNs. For PTO-based ODNs, the most stimulatory motifs in mice show weak activity in human or primate cells. Moreover, sequence specificity is more restricted with PTO-based ODNs compared with native PD-based ODNs (17, 18, 37, 38). Nevertheless, we demonstrated previously that the minimal sequence requirements for activation of hTLR9 are independent of the backbone chemistry (20, 21). The role of hQ364, R348, and Q562 in defining the sequence-specific requirements of hTLR9 also was confirmed for PD-based C-type ODN, which was a weak activator of hTLR9 and C22, failed to activate wt hTLR9, and weakly activated the triple mutant.

With introduced mutations, the response of hTLR9 was also improved for bacterial gDNA. Protection of the host against TLR9 activation by endogenous DNA relies on the methylation of CpG motifs, a lower frequency of CpG motifs, and a high frequency of G-rich inhibitory sequences (9, 12). The stringency for the hTLR9 response is further increased by the minimal sequence-specific requirement for the two appropriately spaced CpG motifs for effective activation of hTLR9. The more stringent specificity of hTLR9 may be due to the differences in the impact of microbial infections. This difference is further apparent in the response to mammalian gDNA, for which the difference in the frequency of the mODN motif versus the hODN motif is amplified by CpG methylation.

The hQ346, R348, and Q562 mutations of hTLR9 to mouse counterparts augmented the activation by methylated and unmethylated mammalian gDNA, with bacterial DNA being an even better activator.

In conclusion, we propose to extend a model of hTLR9 activation by longer ODNs, which takes into account the differences in the immunostimulatory CpG motifs. The previously identified CpG binding site (5) probably accommodates the 5′ CpG motif of the hODN; however, for robust receptor activation, additional sites of the ECD may have to participate to stabilize the signaling TLR9–ODN complex. Although we cannot conclusively state that the second CpG motif binds to the residues identified as having an effect on TLR9 signaling, our results clearly demonstrate their role in the selectivity of stimulatory ODNs. Based on our results, we propose that the additional site acts as an additional regulator for the recognition of agonists and TLR9 activity. Many TLRs are characterized by more than one binding site for the agonists. TLR8 binds uridine and the UG dinucleotide of ssRNA (7), TLR5 binds flagellin at at least two segments across its entire D1 domain (39), two pseudosymmetric binding sites on TLR3 ECD recognize the two sites of the dsRNA duplex (4043), and two sites of TLR4 bind the complex of the coreceptor MD-2 with LPS with a 2:2:2 stoichiometry (44). Improved selectivity and differentiation between microbial and endogenous molecules are the reasons for multipartite recognition by TLRs. Rather than optimizing the receptor for maximal sensitivity to the agonists, the immune receptors have to take into account the selectivity between the associated endogenous and foreign danger patterns, which may be the reason for the differences between the human and mouse receptors.

We thank Ana Kunšek and Katja Leben for technical assistance with ELISA and Rok Gaber and Rok Fink for providing the bacterial strains.

This work was supported by the program and projects of the Slovenian Research Agency and the EN-FIST Centre of Excellence.

The online version of this article contains supplemental material.

Abbreviations used in this article:

BM

bone marrow

BMcDC

BM-derived cDC

Bt-gDNA

DNA from calf thymus

cDC

conventional DC

DC

dendritic cell

ECD

ectodomain

gDNA

genomic DNA

h

human

HA

hemagglutinin

LRR

leucine-rich repeat

m

mouse

minH

minH75

minM

minM80

ODN

oligodeoxyribonucleotide

PD

phosphodiester

PTO

phosphorothioate

rCD2

rat CD2

SEAP

secreted embryonic alkaline phosphate

wt

wild-type.

1
Gilliet
M.
,
Cao
W.
,
Liu
Y.-J.
.
2008
.
Plasmacytoid dendritic cells: sensing nucleic acids in viral infection and autoimmune diseases.
Nat. Rev. Immunol.
8
:
594
606
.
2
Oldenburg
M.
,
Krüger
A.
,
Ferstl
R.
,
Kaufmann
A.
,
Nees
G.
,
Sigmund
A.
,
Bathke
B.
,
Lauterbach
H.
,
Suter
M.
,
Dreher
S.
, et al
.
2012
.
TLR13 recognizes bacterial 23S rRNA devoid of erythromycin resistance-forming modification.
Science
337
:
1111
1115
.
3
Engel
A.
,
Barton
G. M.
.
2010
.
Compartment-specific control of signaling from a DNA-sensing immune receptor.
Sci. Signal.
3
:
pe45
.
4
Watts
C.
,
West
M. A.
,
Zaru
R.
.
2010
.
TLR signalling regulated antigen presentation in dendritic cells.
Curr. Opin. Immunol.
22
:
124
130
.
5
Ohto
U.
,
Shibata
T.
,
Tanji
H.
,
Ishida
H.
,
Krayukhina
E.
,
Uchiyama
S.
,
Miyake
K.
,
Shimizu
T.
.
2015
.
Structural basis of CpG and inhibitory DNA recognition by Toll-like receptor 9.
Nature
520
:
702
705
.
6
Tanji
H.
,
Ohto
U.
,
Shibata
T.
,
Miyake
K.
,
Shimizu
T.
.
2013
.
Structural reorganization of the Toll-like receptor 8 dimer induced by agonistic ligands.
Science
339
:
1426
1429
.
7
Tanji
H.
,
Ohto
U.
,
Shibata
T.
,
Taoka
M.
,
Yamauchi
Y.
,
Isobe
T.
,
Miyake
K.
,
Shimizu
T.
.
2015
.
Toll-like receptor 8 senses degradation products of single-stranded RNA.
Nat. Struct. Mol. Biol.
22
:
109
115
.
8
Chan
M. P.
,
Onji
M.
,
Fukui
R.
,
Kawane
K.
,
Shibata
T.
,
Saitoh
S.
,
Ohto
U.
,
Shimizu
T.
,
Barber
G. N.
,
Miyake
K.
.
2015
.
DNase II-dependent DNA digestion is required for DNA sensing by TLR9.
Nat. Commun.
6
:
5853
.
9
Stacey
K. J.
,
Young
G. R.
,
Clark
F.
,
Sester
D. P.
,
Roberts
T. L.
,
Naik
S.
,
Sweet
M. J.
,
Hume
D. A.
.
2003
.
The molecular basis for the lack of immunostimulatory activity of vertebrate DNA.
J. Immunol.
170
:
3614
3620
.
10
Ewald
S. E.
,
Barton
G. M.
.
2011
.
Nucleic acid sensing Toll-like receptors in autoimmunity.
Curr. Opin. Immunol.
23
:
3
9
.
11
Leadbetter
E. A.
,
Rifkin
I. R.
,
Hohlbaum
A. M.
,
Beaudette
B. C.
,
Shlomchik
M. J.
,
Marshak-Rothstein
A.
.
2002
.
Chromatin-IgG complexes activate B cells by dual engagement of IgM and Toll-like receptors.
Nature
416
:
603
607
.
12
Uccellini
M. B.
,
Busconi
L.
,
Green
N. M.
,
Busto
P.
,
Christensen
S. R.
,
Shlomchik
M. J.
,
Marshak-Rothstein
A.
,
Viglianti
G. A.
.
2008
.
Autoreactive B cells discriminate CpG-rich and CpG-poor DNA and this response is modulated by IFN-alpha.
J. Immunol.
181
:
5875
5884
.
13
Imaeda
A. B.
,
Watanabe
A.
,
Sohail
M. A.
,
Mahmood
S.
,
Mohamadnejad
M.
,
Sutterwala
F. S.
,
Flavell
R. A.
,
Mehal
W. Z.
.
2009
.
Acetaminophen-induced hepatotoxicity in mice is dependent on Tlr9 and the Nalp3 inflammasome.
J. Clin. Invest.
119
:
305
314
.
14
Krieg
A. M.
,
Yi
A. K.
,
Matson
S.
,
Waldschmidt
T. J.
,
Bishop
G. A.
,
Teasdale
R.
,
Koretzky
G. A.
,
Klinman
D. M.
.
1995
.
CpG motifs in bacterial DNA trigger direct B-cell activation.
Nature
374
:
546
549
.
15
Yi
A. K.
,
Chang
M.
,
Peckham
D. W.
,
Krieg
A. M.
,
Ashman
R. F.
.
1998
.
CpG oligodeoxyribonucleotides rescue mature spleen B cells from spontaneous apoptosis and promote cell cycle entry.
J. Immunol.
160
:
5898
5906
.
16
Sen
G.
,
Flora
M.
,
Chattopadhyay
G.
,
Klinman
D. M.
,
Lees
A.
,
Mond
J. J.
,
Snapper
C. M.
.
2004
.
The critical DNA flanking sequences of a CpG oligodeoxynucleotide, but not the 6 base CpG motif, can be replaced with RNA without quantitative or qualitative changes in toll-like receptor 9-mediated activity.
Cell. Immunol.
232
:
64
74
.
17
Hartmann
G.
,
Weeratna
R. D.
,
Ballas
Z. K.
,
Payette
P.
,
Blackwell
S.
,
Suparto
I.
,
Rasmussen
W. L.
,
Waldschmidt
M.
,
Sajuthi
D.
,
Purcell
R. H.
, et al
.
2000
.
Delineation of a CpG phosphorothioate oligodeoxynucleotide for activating primate immune responses in vitro and in vivo.
J. Immunol.
164
:
1617
1624
.
18
Hartmann
G.
,
Krieg
A. M.
.
2000
.
Mechanism and function of a newly identified CpG DNA motif in human primary B cells.
J. Immunol.
164
:
944
953
.
19
Rankin
R.
,
Pontarollo
R.
,
Ioannou
X.
,
Krieg
A. M.
,
Hecker
R.
,
Babiuk
L. A.
,
van Drunen Littel-van den Hurk
S.
.
2001
.
CpG motif identification for veterinary and laboratory species demonstrates that sequence recognition is highly conserved.
Antisense Nucleic Acid Drug Dev.
11
:
333
340
.
20
Pohar
J.
,
Lainšček
D.
,
Fukui
R.
,
Yamamoto
C.
,
Miyake
K.
,
Jerala
R.
,
Benčina
M.
.
2015
.
Species-specific minimal sequence motif for oligodeoxyribonucleotides activating mouse TLR9.
J. Immunol.
195
:
4396
4405
.
21
Pohar
J.
,
Kužnik Krajnik
A.
,
Jerala
R.
,
Benčina
M.
.
2015
.
Minimal sequence requirements for oligodeoxyribonucleotides activating human TLR9.
J. Immunol.
194
:
3901
3908
.
22
Matsumoto
F.
,
Saitoh
S.
,
Fukui
R.
,
Kobayashi
T.
,
Tanimura
N.
,
Konno
K.
,
Kusumoto
Y.
,
Akashi-Takamura
S.
,
Miyake
K.
.
2008
.
Cathepsins are required for toll-like receptor 9 responses.
Biochem. Biophys. Res. Commun.
367
:
693
699
.
23
Roy
A.
,
Kucukural
A.
,
Zhang
Y.
.
2010
.
I-TASSER: a unified platform for automated protein structure and function prediction.
Nat. Protoc.
5
:
725
738
.
24
Zhang
Y.
2008
.
I-TASSER server for protein 3D structure prediction.
BMC Bioinformatics
9
:
40
.
25
Pettersen
E. F.
,
Goddard
T. D.
,
Huang
C. C.
,
Couch
G. S.
,
Greenblatt
D. M.
,
Meng
E. C.
,
Ferrin
T. E.
.
2004
.
UCSF Chimera--a visualization system for exploratory research and analysis.
J. Comput. Chem.
25
:
1605
1612
.
26
Rutz
M.
,
Metzger
J.
,
Gellert
T.
,
Luppa
P.
,
Lipford
G. B.
,
Wagner
H.
,
Bauer
S.
.
2004
.
Toll-like receptor 9 binds single-stranded CpG-DNA in a sequence- and pH-dependent manner.
Eur. J. Immunol.
34
:
2541
2550
.
27
Brown
D. A.
,
Kang
S. H.
,
Gryaznov
S. M.
,
DeDionisio
L.
,
Heidenreich
O.
,
Sullivan
S.
,
Xu
X.
,
Nerenberg
M. I.
.
1994
.
Effect of phosphorothioate modification of oligodeoxynucleotides on specific protein binding.
J. Biol. Chem.
269
:
26801
26805
.
28
Crooke
R. M.
1991
.
In vitro toxicology and pharmacokinetics of antisense oligonucleotides.
Anticancer Drug Des.
6
:
609
646
.
29
Hartmann
G.
,
Battiany
J.
,
Poeck
H.
,
Wagner
M.
,
Kerkmann
M.
,
Lubenow
N.
,
Rothenfusser
S.
,
Endres
S.
.
2003
.
Rational design of new CpG oligonucleotides that combine B cell activation with high IFN-alpha induction in plasmacytoid dendritic cells.
Eur. J. Immunol.
33
:
1633
1641
.
30
Burge
C.
,
Campbell
A. M.
,
Karlin
S.
.
1992
.
Over- and under-representation of short oligonucleotides in DNA sequences.
Proc. Natl. Acad. Sci. USA
89
:
1358
1362
.
31
Jabbari
K.
,
Bernardi
G.
.
2004
.
Cytosine methylation and CpG, TpG (CpA) and TpA frequencies.
Gene
333
:
143
149
.
32
Dalpke
A.
,
Frank
J.
,
Peter
M.
,
Heeg
K.
.
2006
.
Activation of toll-like receptor 9 by DNA from different bacterial species.
Infect. Immun.
74
:
940
946
.
33
Yasuda
K.
,
Richez
C.
,
Uccellini
M. B.
,
Richards
R. J.
,
Bonegio
R. G.
,
Akira
S.
,
Monestier
M.
,
Corley
R. B.
,
Viglianti
G. A.
,
Marshak-Rothstein
A.
,
Rifkin
I. R.
.
2009
.
Requirement for DNA CpG content in TLR9-dependent dendritic cell activation induced by DNA-containing immune complexes.
J. Immunol.
183
:
3109
3117
.
34
Sun
L.
,
Wu
J.
,
Du
F.
,
Chen
X.
,
Chen
Z. J.
.
2013
.
Cyclic GMP-AMP synthase is a cytosolic DNA sensor that activates the type I interferon pathway.
Science
339
:
786
791
.
35
Roberts
T. L.
,
Idris
A.
,
Dunn
J. A.
,
Kelly
G. M.
,
Burnton
C. M.
,
Hodgson
S.
,
Hardy
L. L.
,
Garceau
V.
,
Sweet
M. J.
,
Ross
I. L.
, et al
.
2009
.
HIN-200 proteins regulate caspase activation in response to foreign cytoplasmic DNA.
Science
323
:
1057
1060
.
36
Viglianti
G. A.
,
Lau
C. M.
,
Hanley
T. M.
,
Miko
B. A.
,
Shlomchik
M. J.
,
Marshak-Rothstein
A.
.
2003
.
Activation of autoreactive B cells by CpG dsDNA.
Immunity
19
:
837
847
.
37
Roberts
T. L.
,
Sweet
M. J.
,
Hume
D. A.
,
Stacey
K. J.
.
2005
.
Cutting edge: species-specific TLR9-mediated recognition of CpG and non-CpG phosphorothioate-modified oligonucleotides.
J. Immunol.
174
:
605
608
.
38
Hartmann
G.
,
Weiner
G. J.
,
Krieg
A. M.
.
1999
.
CpG DNA: a potent signal for growth, activation, and maturation of human dendritic cells.
Proc. Natl. Acad. Sci. USA
96
:
9305
9310
.
39
Yoon
S. I.
,
Kurnasov
O.
,
Natarajan
V.
,
Hong
M.
,
Gudkov
A. V.
,
Osterman
A. L.
,
Wilson
I. A.
.
2012
.
Structural basis of TLR5-flagellin recognition and signaling.
Science
335
:
859
864
.
40
Choe
J.
,
Kelker
M. S.
,
Wilson
I. A.
.
2005
.
Crystal structure of human toll-like receptor 3 (TLR3) ectodomain.
Science
309
:
581
585
.
41
Liu
L.
,
Botos
I.
,
Wang
Y.
,
Leonard
J. N.
,
Shiloach
J.
,
Segal
D. M.
,
Davies
D. R.
.
2008
.
Structural basis of toll-like receptor 3 signaling with double-stranded RNA.
Science
320
:
379
381
.
42
Bell
J. K.
,
Botos
I.
,
Hall
P. R.
,
Askins
J.
,
Shiloach
J.
,
Segal
D. M.
,
Davies
D. R.
.
2005
.
The molecular structure of the toll-like receptor 3 ligand-binding domain.
Proc. Natl. Acad. Sci. USA
102
:
10976
10980
.
43
Pirher
N.
,
Ivičak
K.
,
Pohar
J.
,
Benčina
M.
,
Jerala
R.
.
2008
.
A second binding site for double-stranded RNA in TLR3 and consequences for interferon activation.
Nat. Struct. Mol. Biol.
15
:
761
763
.
44
Park
B. S.
,
Song
D. H.
,
Kim
H. M.
,
Choi
B.-S.
,
Lee
H.
,
Lee
J.-O.
.
2009
.
The structural basis of lipopolysaccharide recognition by the TLR4-MD-2 complex.
Nature
458
:
1191
1195
.

The authors have no financial conflicts of interest.

Supplementary data