Identification of Functional Domains of CXCL14 Involved in High-Affinity Binding and Intracellular Transport of CpG DNA Free

The innate immune response is a biological defense mechanism that plays important roles in pathogen recognition, inflammatory cytokine production, and activation of adaptive immune responses (1, 2). Pathogen-associated molecular patterns elicit innate immune responses in dendritic cells (DCs) via pattern recognition receptors (3). TLR9, a representative pattern recognition receptor, localizes in the endosome/lysosome compartment of DCs and activates innate immune reactions by directly binding unmethylated cytosine-phosphate-guanine–containing DNA (CpG DNA), which is primarily derived from microbes (4, 5). Thus, TLR9 mediates immunosurveillance for bacterial and viral infection by detecting unmethylated CpG DNA (4–9).

Synthetic CpG oligodeoxynucleotides (ODNs) are used experimentally as ligands for TLR9 (5, 10) and are classified into three classes based on their structure and function (10). A-class CpG ODNs possess poly-G and palindromic sequences with strong type I IFN–inducing activity in plasmacytoid DCs. B-class CpG ODNs do not contain palindromic sequences and predominantly act on B cells. Finally, C-class CpG ODNs have properties of both A-class and B-class CpG ODNs and contain palindromic sequences at the 3'-end (10). Because TLR9 resides within the endosomes and lysosomes of DCs (11), CpG ODNs must be delivered into these compartments.

Chemokines play a primary role in mobilizing immune cells to guide leukocyte trafficking in inflammation, immune responses, and homeostasis (12). In addition to their typical chemotactic function, several chemokines directly inhibit growth of Gram-negative and Gram-positive bacteria, as well as HIV infection in vitro, by interacting with components of the pathogens (13–16). However, it remains unclear whether this interaction is linked to innate immune responses. Previously, our research group reported that mice lacking the chemokine CXCL14 exhibit weaker inflammatory responses when they become obese (17). Our detailed investigation of CXCL14 functions revealed that CXCL14 directly binds CpG ODN, delivers it into the endosomal/lysosomal compartment of DCs via receptor internalization, and activates TLR9-mediated Th1 cytokine production (18). These findings were consistent with the observation that the Th1 response is stronger in CXCL14 transgenic mice than in wild-type (WT) mice in the collagen-induced arthritis model (19) and in response to transplantation of B16 melanoma cells (20).

We previously reported that CXCL14 binds to CXCR4, the canonical receptor for CXCL12 (21). However, CXCL12 does not carry CpG ODN into DCs at low concentrations (18), and disruption of CXCR4 does not affect the intracellular uptake of CpG ODN/CXCL14 (18). Thus, CXCL12 would be a useful reference for understanding the molecular basis of the CXCL14-mediated CpG ODN transport into bone marrow–derived DCs (BMDCs). In contrast to CXCL12, CXCL16 facilitates incorporation of DNA into DCs and promotes the CpG DNA-mediated secretion of inflammatory cytokines (22). In addition, recent work showed that CXCL4 increases CpG DNA-mediated TLR9 activation (23). An elevated level of CXCL4 in plasma is an early biomarker for systemic sclerosis and positively correlates with serum IFN-α (IFN-α) level, which is associated with the pathogenesis of systemic sclerosis (24). As with CXCL14, CXCL4 forms a stable complex with bacterial or human DNA fragments and activates the TLR9 signaling pathway in mouse plasmacytoid DCs (23). Interestingly, disruption of CXCR3, the canonical receptor for CXCL4, does not affect the CpG DNA/CXCL4-mediated induction of IFN-α (23). Collectively, these results indicate that a second (noncanonical) function of CXC chemokines is transport of CpG DNA to TLR9.

These findings indicated that the interaction between CpG DNA and chemokine is important for enhancing the TLR9 signaling pathway. However, it remains unclear whether formation of the DNA/chemokine complex is sufficient to trigger this reaction to DCs. In addition, CXCL4, CXCL12, and CXCL14 have not been compared under in the same assay conditions. In this study, we evaluated the biological activities of three CXC chemokines and CXCL14-derivatives. The results revealed that CXCL14 has two distinct domains for the binding and intracellular transport of CpG ODN. Our results provide a molecular basis for a noncanonical function of CXC chemokines.

C57BL/6N mice were obtained from Nihon SLC (Hamamatsu, Japan). RAW 264.7 cells were maintained in RPMI-1640 medium (Nacalai Tesque, Kyoto, Japan) containing 10% FBS (Thermo Fisher Scientific, Waltham, MA). BMDCs were prepared from 6-wk-old C57BL/6N mice by culturing bone marrow cells for 10 d in RPMI-1640 medium (Nacalai) containing 10% FBS (Thermo Fisher Scientific), mouse granulocyte–macrophage colony stimulating factor (10 ng/ml; Peprotech, Rocky Hill, NJ, USA), and mouse IL-4 (10 ng/ml; PeproTech). All mice were housed in a specific pathogen–free animal facility under a 12-h light/dark cycle. All experimental procedures were preapproved by the ethical committee of the Tokyo Metropolitan Institute of Medical Science.

Two types of phosphorothioate CpG oligonucleotides, ODN2395 (5'-TCGTCGTTTTCGGCGCGCGCCG-3') and ODN1826 (5'-TCCATGACGTTCCTGACGTT-3'), were purchased from InvivoGen (San Diego, CA) and Eurofin Genomics (Tokyo, Japan), respectively. Both ODNs were 5'-labeled with Cy3 by Eurofin Genomics. Human CXCL12 and human CXCL4 were purchased from BioLegend (San Diego, CA). Propidium iodide (PI) was purchased from Sigma-Aldrich (St. Louis, MA). Antiphosphorylated p38 (no. 9211), and anti-p38 Ab (no. 8690) were purchased from Cell Signaling Technology (Danvers, MA). Polyclonal anti-CXCL14 Ab, which was produced in rabbit and affinity purified from serum, was provided by Takeda Pharmaceutical Company (Osaka, Japan). Normal rabbit IgG was purchased from Wako Chemicals (Osaka, Japan).

All peptide fragments, including C-terminal N-sulfanylethylanilide (SEAlide)–modified peptides, were prepared by standard solid-phase peptide synthesis. CXCL14 and CXCL14-C were chemically synthesized based on the human CXCL14 aa sequence as previously described (25, 26). The synthesis of CXCL14-N was carried out by a selective disulfide bond formation reaction after condensation of CXCL14 (1–28)-SEAlide and CXCL14 (29–50) fragments by native chemical ligation (27). CXCL14 deletion or chimera peptides (CXCL14 [1–47], CXCL14 [1–40], CXCL14 [13–50], and CXCL12/14) and the corresponding biotinylated peptides were synthesized in a similar manner. To avoid misfolding by unpaired cysteine residues, we replaced all cysteine residues in the deletion peptides with alanine by desulfurization (28). Sequence information for the CXCL14 and CXCL14-derived peptides used in this study is provided in Table I. All peptides were HPLC-purified using a Cosmosil 5C18-AR-II column (Nacalai Tesque); each peptide yielded a single peak with consistent retention time. Mass spectra were recorded on a MICROMASS LCT PREMIER (Waters, Milford, MA) or Prominence-I LC-2030, LCMS-2020 (Shimadzu, Kyoto, Japan) to confirm their amino acid sequences. Detailed synthetic methods will be provided upon request.

BMDCs were plated onto 35-mm glass plates (IWAKI, Tokyo, Japan) and incubated for 1 h at 37°C with 30 nM Cy3-ODN2395 with or without CXCL14, CXCL4, or CXCL12 in RPMI-1640 medium containing 20 mM HEPES-NaOH (pH 7.5) and 0.1% fatty acid–free BSA (Sigma-Aldrich). Cells were fixed with 4% paraformaldehyde/PBS for 1 h and stained with DAPI (1 μg/ml; Dojindo, Kumamoto, Japan). Fluorescence images were acquired on a BZ-X700 microscope (Keyence, Tokyo, Japan).

BMDCs were plated at 10⁵ cells per well in 24-well plates (Corning, New York, NY) and incubated for 1 h at 37°C with Cy3-ODN2395 or Cy3-ODN1826 in the presence or absence of the indicated chemokines or CXCL14 derivatives in RPMI-1640 medium containing 20 mM HEPES-NaOH (pH 7.5) and 0.1% fatty acid–free BSA (Sigma-Aldrich). For cell-surface binding assays, BMDCs were trypsinized and incubated for 1 h at 4°C with Cy3-ODN2395 in the presence or absence of CXCL14. To evaluate the effects of chemical inhibitors, BMDCs were pretreated with pertussis toxin (PTX) (200 ng/ml; Sigma-Aldrich), chlorpromazine (10 μg/ml; Sigma-Aldrich), or nystatin (50 μg/ml; Sigma-Aldrich) for 30 min and then incubated for 1 h at 4°C with Cy3-ODN2395 or Cy3-ODN2395/CXCL14 in the presence of the same inhibitor. Cells were trypsinized, stained with allophycocyanin-conjugated mouse anti-CD11c Ab (N418; BioLegend) and PI, and analyzed on an LSRFortessa X-20 (BD Biosciences, San Jose, CA). CD11c⁺PI^– cells were gated for analysis of BMDCs.

BMDCs or RAW 264.7 cells were plated at 10⁵ cells per well in 24-well plates (Corning) and incubated for 6 h at 37°C in RPMI-1640 containing 10% FBS in the presence or absence of ODN2395, and the indicated chemokine or CXCL14 derivative. Culture supernatants were subjected to analyses with ELISA MAX IL-12p40 or TNF-α Kits (BioLegend).

A bicistronic expression vector for Cas9 and guide RNA (gRNA) (pSpCas9 [BB]-2A-GFP; Addgene, Cambridge, MA) was digested with BbsI (New England Biolabs, Ipswich, MA). A pair of oligonucleotides (Table II) for each gRNA, targeting the two loci of the mouse Tlr9 gene (Supplemental Fig. 1A), was annealed and ligated into the linearized vector. RAW 264.7 cells were cotransfected with 2 μg of Cas9/Tlr9–gRNA-1 and -2 expression vectors. Forty-eight hours after transfection, GFP⁺ cells were sorted on an FACSAriaIII (BD Biosciences). Sorted cells were subjected to cell cloning by limiting dilution. Correct genome editing (deletion of most of exon 2) was verified by sequencing of PCR products containing the target sites (Supplemental Fig. 1C). Sequencing data were visualized in 4Peaks (Nucleobytes, Amsterdam, The Netherlands). Primers are listed in Table II.

RAW 264.7 and Tlr9-knockout (KO) RAW 264.7 cells were treated with 300 nM chemokines with or without 100 nM of ODN2395 in RPMI-1640 medium containing 20 mM HEPES-NaOH (pH 7.5) and 0.1% fatty acid–free BSA. After 30 min stimulation, cells were trypsinized and fixed with 2% paraformaldehyde/PBS. The fixed cells were permeabilized with 2% FBS-PBS containing 0.2% NP-40 (Sigma-Aldrich), stained with phospho-p38 Ab (no. 9211, 1:200 dilution; Cell Signaling Technology) and incubated with Alexa Fluor 647–conjugated anti-rabbit IgG (Abcam, Cambridge, U.K.) and CD11b-FITC Ab (M1/70, BioLegend). CD11b⁺ cells were gated to analyze the percentage of phospho-p38⁺ cells.

Peptides and cells were dissolved in SDS sample buffer (250 mM Tris-HCl [pH 8.0], 2% SDS, 5% sucrose, 0.002% bromophenol blue, 100 mM 2-ME [Sigma-Aldrich], 1× complete [Roche Diagnositics], 1× PhosBlock [Roche]). Lysates were separated by SDS-PAGE and electroblotted onto PVDF membranes. Bound proteins were probed with anti-CXCL14 (1 μg/ml), anti–phospho-p38 (no. 9211, 1:500 dilution), or anti-p38 Ab (no. 8690, 1:500 dilution), incubated with peroxidase-conjugate anti-rabbit IgG (GE Healthcare), and visualized by ECL (GE Healthcare). Chemiluminescence signals were detected on an LAS3000 (Fuji Film, Tokyo, Japan).

Modeling of three-dimensional structure the ODN2395/CXCL14 complex was performed using molecular operating environment (MOE) 2019.01 software (Chemical Computing Group ULC, Montreal, Canada). The NMR structure of CXCL14 (Protein Data Bank: 2HDL) (29) was loaded into MOE. The structure of ODN2395 (diphosphate bond sequences) was composed in DNA Builder and loaded into MOE. Protein-DNA docking was applied to the ODN2395/CXCL14 complex. Molecular mechanics–generated Born interaction energy was calculated using Amber10/EHT of MOE. Structural modeling analyses of alanine-substituted CXCL14 peptides were calculated using PyMOL 2.3.2 (Schrodinger, New York, NY). The mutagenesis function of PyMOL was applied to alanine-substituted CXCL14 (1–47) CXCL14 (1–40), and CXCL14 (13–50) peptides.

Pull-down assays were performed as previously described (18). In brief, each biotinylated CXCL14 derivative (1 nmol) was mixed with streptavidin–agarose (Sigma-Aldrich). To that mixture, 100 nM Cy3-ODN2395 with or without unlabeled 10 μM of ODN2395 in 100 μl of binding buffer was added, and the sample was incubated for 1 h at 4°C. To measure Cy3 fluorescence, the agarose beads were subjected to SDS-PAGE. Thereafter, each gel was blotted onto a 0.22-μm PVDF membrane, and biotinylated CXCL14 derivatives were detected by chemiluminescence.

Chemokine or CXCL14-derived peptides (0.5–2 μM) were coated onto H-type multiwell plates (Sumitomo Bakelite, Tokyo, Japan) at 4°C overnight. The coated wells were blocked with 2% BSA (Wako Chemicals) in PBS and incubated with various concentrations of Cy3-ODN2395 (25–800 nM) in binding buffer (50 mM HEPES [pH 7.5], 150 mM NaCl, 1 mM CaCl₂, 1 mM MgCl₂, 1% BSA) for 1 h at room temperature. Each supernatant was collected to measure the amount of free Cy3-ODN. Bound Cy3-ODN was eluted with 100 μl of elution buffer (Tris-HCl [pH 7.5], 200 mM NaCl, 5 mM EDTA, 1% SDS). Cy3 fluorescence was measured using a Varioskan Flash microplate reader (Thermo Fisher Scientific). Averages of triplicated samples were plotted, and analyzed by nonlinear regression using Prism 8 (GraphPad Software, San Diego, CA). For Scatchard plot analyses, the bound/free ratio was then plotted against the concentration of bound Cy3-ODN2395 (nM). The slope of the linear regression line was determined, and K_d was calculated as –1/slope.

FACS and ELISA experiments in the same set of figures were carried out in triplicated wells and replicated from an independent preparation of BMDCs. FACS plot was presented with percentage of Cy3ODN^high fraction of representative experiment. For the CpG DNA uptake assays, statistical analyses were performed by one-way ANOVA and Sidak multiple comparison test relative to control groups. For the ELISA assays, statistical analyses were performed by one-way ANOVA analysis for ODN⁺ groups and Dunnett multiple comparison test relative to control groups. Bar graphs present data as means ± SD with individual value as shown in dot. The aforementioned statistical analyses and calculation of EC₅₀ were performed in Prism 8. The mean fluorescence intensity and percentage of Cy3^high cells were calculated using the FlowJo software (Tomy Digital, Tokyo, Japan).

As we previously reported (18), ODN2395 (C-class) and ODN1826 (B-class) have the highest binding affinities for CXCL14 (K_d = 9.8 nM and 28 nM, respectively). To obtain insight into the CpG ODN/CXC chemokine interaction, we selected ODN2395 and ODN1826 in the following experiments. First, we compared the functions of CXCL4 and CXCL12 with CXCL14. Cy3-labeled ODN2395 (Cy3-ODN2395) alone at a concentration of 30 nM did not increase Cy3 fluorescence in BMDCs (Fig. 1A). By contrast, Cy3-ODN2395 in combination with CXCL14 promoted Cy3 fluorescence in BMDCs (Fig. 1A). CXCL4 also increased Cy3 fluorescence in BMDCs following Cy3-ODN2395 treatment (Fig. 1A). In contrast, at a concentration of 300 nM, but not 100 nM, CXCL12 weakly increased Cy3 fluorescence relative to Cy3-ODN2395 alone (Fig. 1A). These observations suggested that CXCL14 and CXCL4 promote intracellular transport of Cy3-ODN2395 in BMDCs. FACS analysis also indicated that CXCL4 increased incorporation of Cy3-ODN2395, but to a lesser extent than CXCL14 (Fig. 1B, 1C). CXCL12 weakly but not significantly promoted CpG ODN uptake at a high concentration (300 nM) but not at a lower concentration (100 nM) (Fig. 1B, 1C). Incorporation of Cy3-ODN1826 was also promoted by CXCL14 (Fig. 1D, 1E). In contrast to Cy3-ODN2395, intracellular transport of Cy3-ODN1826 was not significantly promoted by CXCL4 and CXCL12 (Fig. 1D, 1E).

Consistent with these DNA transport activities, 100 nM CXCL14 strongly promoted ODN2395-mediated secretion of IL-12p40, which is a downstream target of the TLR9 activation by CpG DNA (4). In contrast, IL-12p40 secretion was induced by ODN2395 plus 300 nM CXCL4 or CXCL12, but at a lower level than by 100 nM CXCL14 (Fig. 2A). EC₅₀ for IL-12p40 induction were determined to be 104 nM for CXCL14, 178 nM for CXCL4, and higher than 1000 nM for CXCL12 (Fig. 2C–E). For ODN1826, 100 nM ODN1826 was capable of stimulating IL-12p40 secretion by itself (Fig. 2B). CXCL14 promoted ODN1826-mediated IL-12p40 secretion, as we previously reported (Fig. 2B). CXCL4 (300 nM) also enhanced ODN1826-mediated IL-12p40 secretion to lesser extent than CXCL14, whereas CXCL12 barely exhibited any such activity (Fig. 2B). The EC₅₀ of CXCL14 for IL-12p40 induction was 218 nM (Fig. 2F). These results indicate that CpG ODN transport activity is shared by three CXC chemokines, whereas the ability to activate TLR9 at the submicromolar range is limited to CXCL14 and CXCL4.

To examine the immediate early response to CpG DNA, we next examined the phosphorylation status of p38, a downstream effecter of TLR9 signaling by using mouse macrophage-derived cell line RAW 264.7 (30). We generated Tlr9-KO RAW 264.7 cells using the CRISPR/Cas9 system (Supplemental Fig. 1A–C). Induction of TNF-α by ODN2395 or ODN2395 + CXCL14 was significantly suppressed in Tlr9-KO RAW 264.7 cells (Supplemental Fig. 1D). Western blotting analysis revealed that ODN2395 + CXCL14 treatment increased p38 phosphorylation (Supplemental Fig. 1E) 30 min after stimulation in WT RAW 264.7 cells. We quantified the ratio of p38-activated cells by FACS (Supplemental Fig. 1F–H). In WT RAW 264.7 cells, ODN2395 + CXCL14 treatment significantly increased p38 phosphorylation (Supplemental Fig. 1F, 1G). By contrast, p38 phosphorylation was not elevated in Tlr9-KO RAW 264.7 cells after stimulation with ODN2395 + CXCL14 (Supplemental Fig. 1F, H). Stimulation with CXCL14, CXCL4, CXCL12, ODN2395, ODN2395 + CXCL4, or ODN2395 + CXCL12 did not significantly increase p38 phosphorylation in WT or Tlr9-KO RAW 264.7 cells (Supplemental Fig. 1F–H). These biochemical results supported the idea that the CpG ODN/CXCL14 complex strongly activates the TLR9 signaling pathway with rapid kinetics.

To determine which pathway is involved in the internalization of CpG ODN/CXCL14 complex, we treated BMDCs with Cy3-ODN2395 or Cy3-ODN2395/CXCL14 in the presence of chemical inhibitors. PTX, an inhibitor of G_i-coupled GPCRs including canonical chemokine receptors, did not affect Cy3-ODN2395 incorporation (Fig. 3A, 3B). Cellular uptake of Cy3-ODN2395 was suppressed by chlorpromazine and nystatin, which are inhibitors of clathrin and caveolin-dependent endocytosis, respectively (Fig. 3A, 3B). Importantly, internalization of the Cy3-ODN2395/CXCL14 complex or Cy3-ODN2395 (300 nM) was greatly suppressed by chlorpromazine, but not by nystatin (Fig. 3). These observations suggested that intracellular uptake of the Cy3-ODN2395/CXCL14 complex occurs via clathrin-dependent endocytosis.

To determine whether the increase in internalization of CpG ODN by CXCL14 was mediated by cell-surface receptors, we used an anti-CXCL14 rabbit polyclonal Ab generated by our group. This Ab, which recognized CXCL14, but not CXCL12 or CXCL4 (Supplemental Fig. 2A), blocked internalization of the Cy3-ODN2395/CXCL14 complex in a dose-dependent manner (Fig. 3C, 3D), whereas normal rabbit IgG did not (Fig. 3C, 3D). Because Cy3-ODN2395 uptake without CXCL14 was not blocked by this Ab (Supplemental Fig. 2B), this effect was specific to Cy3-ODN2395/CXCL14 complex. Finally, anti-CXCL14 Ab, but not normal rabbit IgG, blocked the cell-surface binding of the Cy3-ODN2395/CXCL14 complex (Fig. 3E), but not the interaction between ODN2395 and CXCL14 (Supplemental Fig. 2C). These results indicated that unidentified CXCL14 receptor(s) on BMDCs are involved in endocytosis of the CpG ODN/CXCL14 complex.

CXCL14 has two characteristic structural domains, an antiparallel β-sheet in the N-terminal region (aa residues 1–49) and an α-helix in the C-terminal region (residues 50–77) (25). To determine which domain is required for CpG ODN-mediated TLR9 activation, we synthesized an N-terminal peptide including residues 1–50 (designated as CXCL14-N) and a C-terminal peptide including residues 50–77 (CXCL14-C) (Table I). CXCL14-N promoted incorporation of Cy3-ODN2395 by BMDCs, although to a lesser extent than WT CXCL14 (Fig. 4A, 4B). CXCL14-N slightly increased the uptake of Cy3-ODN1826. However, this effect was not statistically significant relative to Cy3-ODN1826 alone (Fig. 4C, 4D). In contrast, CXCL14-C had no CpG ODN transport activity for Cy3-ODN2395 and Cy3-ODN1826 (Fig. 4A–D). As expected, CXCL14-N–stimulated IL-12p40 secretion occurred in combination with either ODN2395 or ODN1826, but the activity was weaker than that of WT-CXCL14 (Fig. 4E, 4F). CXCL14-C did not promote CpG ODN-mediated IL-12p40 secretion (Fig. 4E, 4F). Taken together, these observations suggest that CXCL14-N is sufficient for cytokine induction, and that the C-terminal domain cooperates with CXCL14-N to strengthen its activity.

CXCL14 contains two loop structures at N-terminal residues 1–12 and internal residues 41–47, as well as a short β-sheet structure encompassing residues 48–50 before C-terminal α-helix (29). To determine whether these loop domains are required for the binding and intracellular transport of CpG ODN, we synthesized a series of deletion peptides: CXCL14 (1–47), CXCL14 (1–40), and CXCL14 (13–50). CXCL14 has four cysteine residues that form two pairs of disulfide bonds (Cys³-Cys²⁹, Cys⁵-Cys⁵⁰). To avoid misfolding because of unpaired cysteine residues, we replaced all cysteine residues in the deletion peptides with alanine by desulfurization (28) (Table I). Modeling analysis using PyMOL predicted that these alanine substitutions would not alter overall secondary structure (Fig. 5A–C).

Incorporation of Cy3-ODN2395 was promoted by CXCL14 (1–47) to a similar extent as CXCL14-N (Fig. 5D, 5E). Moreover, CXCL14 (1–47) promoted ODN2395-mediated IL-12p40 secretion to a similar extent as CXCL14-N (Fig. 5F). These observations suggest that the two disulfide bonds and residues 48–50 of CXCL14 are dispensable for its activity. In contrast, the ODN2395 transport activities of CXCL14 (1–40) and CXCL14 (13–50) were lower than that of CXCL14-N (Fig. 5D, 5E). CXCL14 (1–40) and CXCL14 (13–50) did not promote ODN2395-mediated IL-12p40 secretion (Fig. 5F). Next, we carried out pull-down assays using biotinylated CXCL14-derived peptides. Both CXCL14 (1–47) and CXCL14 (1–40), but not CXCL14 (13–50), specifically bound Cy3-ODN2395 (Fig. 5G). These observations indicated that residues 1–12 of CXCL14 are essential for ODN2395 binding, whereas 41–47 are involved in intracellular transport.

To further explore the role of the N-terminal loop of CXCL14, we synthesized a chimeric peptide containing residues 1–22 (N-terminal loop) of CXCL12 followed by residues 13–47 of CXCL14 (Fig. 6A, Table I). This CXCL12/14 chimera exhibited ODN2395 binding capacity (Fig. 6B), suggesting that the N-terminal loop of CXCL14 can be replaced by that of CXCL12 to compensate for CpG ODN binding. This observation also supported the idea that the N-terminal loop of CXCL12 is capable of generating a high-affinity ODN binding site. However, the CXCL12/14 chimera did not restore the defect of CXCL14 (13–50) in ODN2395 transport (Fig. 6C, 6D). Similarly, the CXCL12/14 chimera did not promote ODN2395-mediated IL-12p40 secretion (Fig. 6E). These findings imply that the N-terminal loop of CXCL14 plays important roles not only in CpG ODN binding, but also in internalization of CpG ODN via cell-surface receptors.

To determine the binding affinity of CXCL14 and other chemokines for CpG ODN, we performed plate-based binding assays. For these experiments, we coated ELISA plates with each chemokine and added various concentrations of Cy3-ODN2395. Cy3-ODN2395 bound to CXCL12 or CXCL14-coated plates, but not to noncoated plates, and binding increased in a dose-dependent manner up to a concentration of 400 nM of CXCL14 or CXCL12 (Fig. 7A). Although Cy3-ODN2395 bound to CXCL4-coated plates less strongly than to CXCL14- or CXCL12-coated plates, the binding kinetics were similar to those of CXCL14 or CXCL12 (Fig. 7A). Scatchard plot analyses based on this system revealed that CXCL14 binds to Cy3-ODN2395 with K_d = 37 nM (Supplemental Fig. 3A). CXCL4 had a lower binding affinity for Cy3-ODN2395 (K_d = 142 nM) (Supplemental Fig. 3B), consistent with its weaker ODN2395 transport activity, relative to CXCL14 (Fig. 1A–C). Interestingly, CXCL12 bound to Cy3-ODN2395 with high affinity (K_d = 18 nM) (Supplemental Fig. 3C) despite its very low ODN2395 transport activity (Fig. 1A–C).

We also observed saturable binding of Cy3-ODN2395 on plates coated with CXCL14-N and CXCL14-C peptides (Fig. 7B). Scatchard plot analyses showed that CXCL14-N and CXCL14-C bound to Cy3-ODN2395 with K_d values of 96 nM and 72 nM, respectively (Supplemental Fig. 3D, 3E). These K_d values were larger than that of WT-CXCL14, suggesting that the high-affinity ODN2395 binding of WT-CXCL14 was supported by the cooperative action of the N- and C-terminal regions. We observed similar saturable binding when using CXCL14 (1–47), CXCL14 (1–40), or CXCL12/14 chimera (Fig. 7C). Consistent with the results of pull-down experiments, Cy3-ODN2395 bound to the CXCL14 (13–50)–coated plate weakly and in a nonsaturable manner (Fig. 7C), suggesting low-affinity binding. Scatchard plot analyses revealed that CXCL14 (1–47), CXCL14 (1–40), and CXCL12/14 chimera bound to Cy3-ODN2395 with K_d values of 76, 81, and 102 nM, respectively (Supplemental Fig. 3F–H). We noticed that CXCL14-C, CXCL14 (1–40), and CXCL12/14 chimera possessed the higher binding capacity for Cy3-ODN2395 than CXCL4, but these peptides lacked ODN2395 transport activity. Together with the results obtained with CXCL12, these observations indicate that CpG ODN binding capacity is necessary but not sufficient for intracellular transport.

Finally, we performed structural modeling analysis for the ODN2395/CXCL14 complex using a protein–DNA docking program in MOE. The best-fit model with the highest score is shown in (Fig. 7D. Potential energy calculated by this simulation was -226 kcal/mol. The Site Finder module of the MOE identified five possible ligand binding sites in the modeled ODN2395/CXCL14 complex. Arginine (R)7 and lysine (K)8 in the N-terminal loop of CXCL14 recognize a phosphate bond between cytosine (C)2 and guanine (G)3 of ODN2395. K11 in the N-terminal loop of CXCL14 also recognizes guanine base of G3 (Fig. 7D). Serine (S)57 and R60 in the C-terminal α-helix recognize a phosphate bond between G3 and thymine (T)4 of ODN2395 (Fig. 7D). These observations support the cooperative action of CXCL14-N and C in the high-affinity binding of CXCL14 to ODN2395.

In this study, we investigated the CpG ODN binding and transport activity of CXCL4, CXCL12, CXCL14, and their derivatives. CXCL12, CXCL14-C, CXCL14 (1–40), CXCL14 (13–50), and a CXCL12/14 chimera bound CpG ODN, but did not transport CpG ODN into BMDCs. Consequently, these peptides were unable to promote IL-12p40 secretion. These data imply that for a carrier peptide to be fully active, CpG ODN binding capacity is necessary, but not sufficient, for intracellular transport and subsequent cytokine induction.

The high-affinity interaction between CXCL14 and CpG ODN is supported by cooperative action of the N-terminal loop structure and the C-terminal α-helix. Because N-terminal amino acids 1–12 of CXCL14 contain four Lys residues and one Arg residue, an electrostatic interaction could occur between CXCL14 and CpG ODN. Structural modeling data revealed that the potential sites of interaction between CXCL14 and ODN2395 are R7, K8, K11, S57, and R60. This model is consistent with our binding data for the CXCL14 deletion peptides. Our simulation model also indicated that a binding pocket for phosphate bonds of ODN2395 is formed by interaction with the N-terminal loop and C-terminal α-helix of CXCL14. The N-terminal loop interacts with the C-terminal α-helix via hydrophobic residues to create a close contact between them, which may be related to the efficient interaction between CpG ODN and CXCL14.

Previously, we showed that CXCL14 promotes binding of ODN2395 to surface receptors of BMDCs. In this study, we showed that anti-CXCL14 Ab blocked cell-surface binding and internalization of the CpG ODN/CXCL14 complex. Because this Ab did not interfere with binding between CpG ODN and CXCL14, it is likely that as-yet-unidentified cell-surface receptor(s) for CXCL14 are involved in the CpG ODN/CXCL14 internalization. On a related note, BMDCs lacking CXCR4, a receptor for CXCL14, produce IL-12p40 upon stimulation with CpG ODN and CXCL14 (18, 21). Similarly, disruption of CXCR3 does not affect CpG DNA/CXCL4–mediated production of IFN-α (23). Interestingly, the biological activities of CXCL4 are not always mediated by CXCR3 signaling, as the plasma concentration of CXCL4 is too high to activate G protein–coupled receptors (31). Consistent with this, the activity of CpG DNA/CXCL4 is independent of CXCR3. In addition, we showed that PTX, an inhibitor of canonical CXC chemokine receptors, did not affect intracellular uptake of the CpG ODN/CXCL14 complex. Therefore, conventional CXC chemokine receptors might not be used in CpG ODN transport. Because internalization of the Cy3-ODN2395/CXCL14 complex was suppressed by treatment with chlorpromazine, a clathrin-dependent endocytosis inhibitor, the candidate receptor(s) must be recycled. It remains to be determined whether the CpG ODN/CXCL14 and CpG ODN/CXCL4 complexes use the same receptor molecule(s) for TLR9 activation. Proteoglycans bind several chemokines via their sugar moieties, glycosaminoglycans (32), and they could be the common receptors shared by CXCL14 and CXCL4.

Another property of the cationic peptide/CpG ODN complex is formation of a self-assembled crystalline structure that activates TLR9. Cationic bactericidal peptides such as LL37 and β-defensin facilitate incorporation of CpG ODN by DCs and promote CpG ODN-mediated TLR9 activation (33–35). Both these peptides and CXCL4 form a liquid crystalline complex with DNA (23, 36, 37). Small-angle x-ray scattering analyses revealed that this liquid crystalline structure contains a columnar DNA bundle with inter-DNA spacing of 3.5–4 nm (23, 37). This spacing is suitable for multimerization of TLR9, which amplifies its signaling (37). Interestingly, CXCL14 (1–40) has high affinity for ODN2395, but despite having binding capacity, it is unable to transport the DNA. In the case of the CXCL14/CpG ODN complex, a loop structure between the second and third β-sheets may be critical for the formation of the liquid crystalline structure.

In this study, we demonstrated that CXCL14 and CXCL4 are functional CpG ODN transporters involved in fully activating the TLR9 signaling pathway. Moreover, we uncovered the molecular basis for the specific interaction between CXCL14 and CpG ODN. Formation of the CXCL14/CpG DNA complex could be involved in progression of, or immune defense against, collagen-induced arthritis, obesity-associated diabetes, and carcinogenesis.

We thank Dr. Shinji Amari for helpful instruction on MOE operation, and Takeda Pharmaceutical for providing a polyclonal Ab against CXCL14. K.T. thanks the members of the laboratory group for advice and scientific suggestions.

A.S., A.O., T.H., and K.T. have a Japanese patent pending (2017-200914). The other authors have no financial conflicts of interest.