Toll-like receptor (TLR) 3 immune modulation by unformulated small interfering RNA or DNA and the role of CD14 (in TLR-mediated effects)

Abstract

The Toll-like receptors (TLRs) 3, 7, 8 and 9 stimulate innate immune responses upon recognizing pathogen-derived nucleic acids. TLR3 is located on the cell surface and in cellular endosomes and recognizes double-stranded viral RNA or the synthetic mimic poly rI:rC. Recently, unformulated small interfering RNA (siRNA) has been reported as ligand for surface-expressed murine TLR3. Blockage of TLR3 is achieved by single-stranded DNA. We confirm and expand the observation that poly rI:rC-mediated TLR3 immune activation is blocked in a sequence-, length-, backbone- and CpG-dependent manner. However, human TLR3 is not activated by siRNA, which may be the result of differences in the amino acid composition of the TLR3 loop 1 of mice and humans. Although CD14 was previously described as a co-receptor for murine TLR3 and other nucleic acid-recognizing TLRs, human CD14 acts only as co-receptor to human TLR9, but not TLR3, TLR7 or TLR8. We show that CD14 up-regulates the TLR9 immune response of A, B and C-class oligodeoxynucleotides but down-regulates the phosphoro-diester version of B-class oligodeoxynucleotides.

Introduction

Toll-like receptors (TLRs) play an essential role in pathogen recognition and innate immune responses in mammals. In humans, 10 different TLRs recognize a variety of pathogen-associated molecular patterns from bacteria, viruses and fungi.¹ TLR7, TLR8 and TLR9 belong to the same subfamily of TLRs based on their genomic structure, sequence similarities and endosomal localization,² and they respond to certain nucleic acids or derivatives. In addition, TLR7 and TLR8 are stimulated by small molecules such as Resiquimod.³ Single-stranded, GU-rich RNA from viruses or synthetic single-stranded oligoribonucleotides (ORN) have been described as TLR7 and TLR8 ligands, whereas TLR9 recognizes non-methylated, CpG-containing DNA of bacterial or viral origin, or synthetic CpG oligodeoxynucleotides (ODN).^3–9

TLR3 is located on the cell surface and in intracellular endolysosomal compartments.^10–14 TLR3 recognizes double-stranded viral RNA or the synthetic mimic poly rI:rC.^15,16 Recently mRNA and small interfering RNA (siRNA) in the presence or absence of delivery systems were described as effective TLR3 ligands.^12,17–19 Whereas mRNA and formulated siRNA activate endosomal TLR3 in human and murine models in vivo and in vitro,^17,19–21 stimulation of TLR3 located on the cell surface in response to unformulated siRNAs was only shown for murine TLR3.^12,22,23

RNA interference (RNAi) is a conserved mechanism in plants and animals to block gene expression by sequence-specific degradation of mRNA. RNAi results in the formation of siRNA duplexes that consist of 21–28 nucleotides with 3′-overhangs of two nucleotides at either terminus or with blunt ends.^24–26 Highly specific silencing of target genes makes siRNA a promising tool in drug development. Several non-specific effects can, however, complicate the use of RNAi. One of the most prevalent non-specific effects is the triggering of the innate immune system in mammals. Recently Kleinman et al.¹² showed that only immune activation via surface TLR3, but not RNAi, leads to suppression of blinding choroidal neovascularization by non-specific siRNAs in a murine disease model.

CD14 is another well-known pattern-recognition receptor in the innate immune system. It was originally identified as a lipopolysaccharide (LPS) receptor that is attached to the cell surface by a glycosylphosphatidylinositol anchor.²⁷ Later, it was shown to function as an essential component of the TLR4/MD2 complex in LPS signalling.²⁸ CD14 was recently identified also as a co-receptor for other TLRs, namely murine TLR3, TLR7 and TLR9.^29–31 Enhancement of murine TLR3 signalling, as well as murine TLR7 and TLR9 signalling and ligand uptake, suggested a dual function for CD14 in nucleic acid recognition by murine TLRs.

This study is structured in three different parts analysing the human TLR3 immune modulation as well as the role of human CD14 as co-receptor to TLR3, TLR7, TLR8 or TLR9. First, we analysed the expression pattern and the immune response of surface-expressed TLR3 by unformulated siRNA. We show that TLR3 mRNA is expressed in several human cell lines that differ in their endosomal and surface localization of TLR3. However, in contrast to murine TLR3 none of these human cell lines were responsive to unformulated, naked siRNA. Moreover, expression of a functional human chimeric surface TLR3–4 receptor did not result in TLR3-mediated signalling by naked siRNA. These data suggest that the postulated hypothesis of unformulated siRNA as surface TLR3 ligand may not apply to human cells.^12,22 Second, we further analysed the inhibition of human TLR3 activation by single-stranded DNA. We demonstrated that single-stranded DNA blocks TLR3 signalling in a CpG- and sequence-independent but length- and backbone-dependent manner. Third, additional experiments concentrated on the role of human CD14 as potential co-receptor for human TLR3, TLR7, TLR8 and TLR9. In contrast to previous data demonstrating murine CD14 acting as a specific co-receptor to TLR3, TLR7 and TLR9,^29,31 in our hands human CD14 did not appear to function as co-receptor for human TLR3, TLR7 or TLR8, but for human TLR9. Interestingly, we observed that differences in CD14 regulation of TLR9 signalling are observed for different CpG ODN classes, the A, B or C classes.

Materials and methods

Reagents

ORN, CpG ODN and siRNA were provided by Coley Pharmaceutical GmbH (Düsseldorf, Germany). Poly rI:rC was purchased from GE Healthcare Europe GmbH (Freiburg; Germany). All substances were controlled for identity and purity by Coley Pharmaceutical GmbH and had undetectable endotoxin levels (< 0·1 EU/ml) measured by the Limulus assay (BioWhittaker, Verviers, Belgium). Oligonucleotides and poly rI:rC were suspended in sterile endotoxin-free Tris–EDTA (CpG ODN) (Sigma, St. Louis, MO) or in DNAse and RNAse-free water (ORN) (Life Technologies, Eggenstein, Germany), and were stored and handled under aseptic conditions to prevent contamination. N-[1-(2,3-Dioleoyloxy)propyl]-N,N,N-trimethylammonium methyl-sulphate (DOTAP) was obtained from Roche (Mannheim, Germany) and Dharmafect from Thermo Fisher Scientific (Epsom, UK). Soluble CD14 (sCD14) was purchased from Biometec GmbH (Greifswald, Germany).

Generation of a TLR3–4 chimeric construct

The TLR3 ectodomain³² was amplified from THP-1 DNA using the primers TLR3-HindIII-fwd (5′-CCC-AAGCTT-G-ATCATGAGACAGACTTTGCC-3′) and TLR3-AsnI-rev (5′-TAGC-ATTAAT-AGTTCAAAGGGGGCACTGTC-3′). The transmembrane and cytoplasmic domain of human TLR4 was identified from murine TLR4³³ via multiple sequence alignment (ClustalW: http://www.ebi.ac.uk/Tools/msa/clustalw2/) and amplified from THP-1 DNA using the primers TLR4-AsnI-fwd (5′-TAGC-ATTAAT-ATGAATAAGACCATCATTGGT-3′) and TLR4-NotI-rev (5′-ATAAGAAT-GCGGCCGC-TCTGAGTCGTCTCCAGAAbottom 96-well plates and incubated for 3 days to reach 70–80% confluence at 37°/5% CO₂ before endosomal or surface staining for TLR3. Cells were scraped and re-suspended by multiple pipetting in 100 μl/well of 1× PBS. Surface staining was performed with a 1/10 dilution of phycoerythrin-conjugated TLR3.7 (TLR3.7-PE) (Tebu-Bio, Offenbach, Germany), 1/10 dilution of IgG-PE (Tebu-Bio) or nothing for 1 hr in darkness at 4°. For intracellular staining, cells were fixed and permeabilized with Intraprep (Beckman Coulter; Krefeld, Germany) according to the manufacturer's instructions. Cells were stained with 1/10 dilution of TLR3.7-PE (Tebu-Bio) or 1/10 dilution of IgG-PE (Tebu-Bio) for 1 hr in Intraprep II. Surface and intracellular samples were washed twice in 1× PBS, re-suspended in 1% formaldehyde/1× PBS and analysed on a FACSCalibur flow cytometer (Becton Dickinson) by Cellquest Pro (BD Biosciences, Heidelberg, Germany) with a minimum of 10 000 events.

Immunofluorescence microscopy

Cells were grown on 10% poly l-lysine-coated coverslips located in 24-well plates up to 60–70% confluence at 37°/5% CO₂ before endosomal or surface staining of TLR3. Surface staining was performed with 1/10 dilution of TLR3.7-PE (Tebu-Bio), 1/10 dilution of IgG-PE (Tebu-Bio) or nothing for 1 hr in darkness at 4°. For intracellular staining, cells were fixed and permeabilized with Intraprep) according to the manufacturer's instructions. Cells were stained in a 1/10 dilution with TLR3.7-PE, IgG-PE or nothing for 1 hr in Intraprep II. Cells were washed twice in 1× PBS, dried at room temperature and covered with ProLong Gold anti-fade reagent with DAPI (Molecular Probes; Invitrogen, Life Technologies GmbH, Darmstadt, Germany). Fluorescence microscopy was performed with 40× enlargement on an Olympus BX41.

Results

TLR3 surface expression observed in specific human cell lines

In this first part of our study we investigated the immune response of surface-expressed human TLR3 by unformulated siRNA. Immune activation via surface TLR3 by naked siRNA was analysed in several cell lines as well as in transiently transfected TLR3-expressing cells. Activation of innate immune responses by siRNA is discussed in the literature as occurring via endosomal pathways involving TLR3, TLR7 or TLR8 or cytoplasmic pathways via retinoic acid-inducible gene-I (RIG-I), melanoma-differentiation-associated gene 5 (Mda-5) or protein kinase R (PKR). To investigate the siRNA-mediated TLR3 immune response in human cell lines and to control for potential cross-reaction via TLR7 or TLR8, RIG-I, Mda-5 or PKR, several human endothelial, tumorigenic and hepatocellular cell lines and primary human hepatocytes were characterized for their expression of these receptors using the LightCycler technology (Fig. 1). Whereas TLR7 and TLR8 expression was only observed in a few cell lines (Ramos, Raji and RPMI), TLR3 mRNA was detected in most of these cell lines and in primary cells (A549, Alexander, APRE-19, DYT2, HEK-293, HeLa, hepatocytes, Huh7, Huh7.5, HUVEC, HDMEC, JIMT-1, MB-175, MB-468, N87, Ramos, Raji, RL-95-2, SKBR3, SK-Hep and T98G). However, most of these cell lines also express RIG-I, Mda-5 and PKR mRNA, which may complicate studies on the immune-activating effects of siRNA.

Figure 1.

Cell lines differentially express Toll-like receptors (TLRs), retinoic acid-inducible gene I (RIG-I) -like receptors, protein kinase R (PKR) and CD14. Three to six individual samples of different human cell lines were used for RNA isolation, cDNA synthesis and LightCycler analysis. LightCylcer was performed in duplicates for each sample using Search-LC GmbH or self-made kits (PKR, RIG-I). Data represent absolute expression normalized to 1000 copies of peptidylprolyl isomerase B/cyclophilin B (PPIB). Mean of four independent experiments (n = 4).

The TLR3-positive cell lines were characterized by FACS and fluorescent microscopy for endosomal and cell surface TLR3 expression (Fig. 2). So far, several commercial antibodies exist for TLR3 staining. However, only the PE-conjugated anti-human monoclonal antibody TLR3.7,³⁶ recognizing the full-length TLR3 protein, was found to be specific and did not cross-react with cell lines lacking TLR3 mRNA expression and without response to poly rI:rC such as U937, THP-1 or PC3 (data not shown and Fig. 3). As shown in Fig. 2(a,b), the TLR3.7 antibody revealed, similar to the mRNA data, endosomal and/or surface TLR3 protein expression in HUVEC, APRE-19 and RL-95-2 cell lines, but not in U937 cells. Primary HUVEC cells, eye-tumorigenic APRE-19 and tumorigenic RL-95-2 cell lines showed strong TLR3 surface expression. Intracellular TLR3 was only detected for cell lines APRE-19 and RL-95-2 but not for HUVEC cells. Fluorescent microscopy demonstrated further that these cell lines contain TLR3 on the cell surface and in intracellular compartments (data not shown and Fig. 2c,d for RL-95-2 cells). FACS analysis and the calculation of mean fluorescent intensities were performed for each TLR3 mRNA-positive cell line (Fig. 2e,f). As each cell line showed different levels of autofluorescence (data not shown), the fold induction of the mean fluorescence above medium is given. Nearly all cells expressed (in the range of a twofold deviation) similar levels of endosomal TLR3, except for A549 (fivefold higher expression) and for APRE-19 (15-fold higher expression) (Fig. 2e). Interestingly, only A549, APRE-19, HDMEC, HUVEC, JIMT1 and N87 demonstrated strong TLR3 surface protein expression and appeared to be good candidates to further investigate TLR3 surface immune activation by unformulated siRNA. In summary, we identified several cell lines that differed in their TLR3 endosomal and surface protein expression.

Figure 2.

Cell lines expressing Toll-like receptor 3 (TLR3) demonstrate intracellular and cell surface localization. Cells were incubated for 3 days before surface staining (a, c) or intracellular staining (b, d) with phycoerythrin-conjugated TLR3.7 (TLR3.7-PE) or IgG-PE, or were left untreated. IgG-PE-positive or TLR3.7-PE-positive cells were gated as histogram plots in relation to side scatter of each cell line (a, b) using CellQuest or shown by fluorescence microscopy (c, d). Mean fluorescent intensity (FI) of each histogram plot was calculated for media, IgG-PE and TLR3.7-PE staining. Data for IgG-PE and TLR3.7-PE are shown as fold induction above media (c, d) for intracellular (e) and surface (f) staining. The data represent the mean of six independent experiments (n = 6).

Figure 3.

Unformulated small interfering RNA (siRNA) does not induce cytokine release from human cell lines expressing Toll-like receptor 3 (TLR3) except for RL-95-2. TLR3-expressing cell lines were trypsinized, treated with trypsin-neutralization solution and incubated at 37°/5% CO₂ for 2 days in serum-free AIMV media. Cell stimulation was performed with 100 μg/ml poly rI:rC, 50 μg/ml DOTAP, 5 μl Dharmafect, media or the indicated concentrations of unformulated, DOTAP-formulated (50 μg/ml start concentration) or Dharmafect-formulated (5 μl start concentration) siRNA si-003. After 20 hr, supernatants were harvested and interleukin-6 (IL-6) (A549, Hela, HUVEC, HDMEC, RL-95-2, SK-Hep, T98G) or IL-8 (APRE-19) was measured by ELISA. Data represent mean out of four independent experiments (n = 4). n.d. = not done.

Unformulated siRNA is not capable of stimulating a human TLR3-mediated immune response in vitro

A549, APRE-19, HeLa, HUVEC, HDMEC, RL-95-2, SK-Hep and T98G cell lines or primary cells were selected to test for TLR3-mediated immune activation by poly rI:rC, unformulated siRNA, DOTAP-formulated siRNA or Dharmafect-formulated siRNA (Fig. 3). DOTAP delivers single-stranded DNA, single-stranded RNA and double-stranded RNA to the endosomal compartments,^5,37–39 whereas Dharmafect mainly results in cytoplasmic delivery (refs 7, 40 and manufacturer's protocols). Cell lines were characterized by LightCycler and Luminex for the production of several cytokines and chemokines such as IL-6, IL-8, interferon-β (IFN-β), IL-12 and IFN-γ. The IL-6 and IL-8 were selected as the most sensitive TLR3-mediated cytokines for further experiments (data not shown).

Unformulated siRNA is degraded in serum-containing media within 5 min, whereas in serum-free media higher stability of up to a few hours was observed (data not shown). Therefore, all cell lines and primary cells were grown in serum-free media. HeLa, SK-Hep and T98G cell lines as well as primary HUVEC and HDMEC cells showed an immune response to poly rI:rC, but did not respond to unformulated or encapsulated siRNA. CD14 is critical as a co-receptor for TLR3 activation,³¹ but we were able to exclude a potential influence of CD14 as TLR3 co-receptor because SK-Hep cells express CD14 mRNA (Fig. 1) but do not respond to unformulated siRNA. The observed immune response induced by DOTAP-delivered or Dharmafect-delivered siRNA in HUVEC and HDMEC primary cells appeared to be unspecific, because the delivery systems themselves caused a similar cytokine release. In contrast, an IL-6 immune response to siRNA delivered by Dharmafect was demonstrated for A549, APRE and RL-95-2 cells (Fig. 3). Besides TLR3 and RIG-I those cell lines also express Mda-5 and PKR (Fig. 1), which have also been described as receptors for formulated siRNA.^41–43 Therefore, an immune response to siRNA mediated by RIG-I, Mda-5 or PKR cannot be excluded. Interestingly, from all the tested cell lines only RL-95-2 cells showed low levels of immune response when stimulated by unformulated siRNA. However, because single-stranded GU-rich RNA R-0006 and its negative control R-1263⁶ showed a similar immune response (data not shown), a potential unspecific RNA-based effect could be assumed. In summary, our data argue against an induction of immune effects by naked siRNA via human TLR3.

As shown in Fig. 1, human cell lines and primary cells express TLR3 and contain mRNA for other siRNA recognizing receptors like RIG-I, PKR or Mda-5. To exclude cross-reactivity between these nucleic acid-recognizing receptors and to analyse for TLR3 cell surface recognition of siRNA we generated a human chimeric TLR3–TLR4 construct with a TLR3 ectodomain and a TLR4 transmembrane and cytosolic domain for over-expression in HEK-293 cells. TLR4 is exclusively expressed on the cell surface^44–47 and has already been used to generate chimeric constructs with murine TLR3, TLR7 or TLR9 ectodomains.^{33,45,48–50} The human TLR3 ectodomain was constructed according to de Bouteiller et al.³² and the murine TLR4 transmembrane and cytosolic domain⁴⁸ was used to identify the correct human domain (Fig. 4a).

Figure 4.

Small interfering RNA (siRNA) does not activate Toll-like receptor 3 (TLR3) when expressed on the cell surface. (a) A human TLR3–4 chimeric receptor was generated by using the TLR3 ectodomain and TLR4 transmemembrane domain. (b–d) MD2-CD14-HEK-293 cells were incubated overnight before transient transfection with human TLR3, human TLR3–4 or an empty vector. After 2 days, cells were stained for surface (b and d) or intracellular (c) TLR3 with phycoerythrin-conjugated TLR3.7 (TLR3.7-PE) or IgG-PE. Positive IgG-PE or TLR3.7-PE cells were gated by FACS analysis as histogram plots in relation to the side scatter of each transfection (b and c) or as immunofluorescence images (d). (e) MD2-CD14-HEK-293 cells were incubated overnight before transient transfection with human TLR3 or human TLR3-4 and a nuclear factor-κB (NF-κB) -luciferase read-out. Following 1-day culture, cells were activated with the indicated concentrations of poly rI:rC, unformulated siRNA, DOTAP (50 μg/ml: 1/3 dilution) formulated siRNA, HiPerfect (10 μl/well: one-third dilution) formulated siRNA, 200 ng/ml lipopolysaccharide (LPS), 10 ng/ml tumour necrosis factor-α or left untreated. NF-κB activation was measured by assaying luciferase activity. Results are given as fold induction above background (medium). Data show the mean of four independent experiments (n = 4).

FACS analysis as well as fluorescent microscopy of MD2-CD14-HEK-293 cells transiently transfected with human TLR3, human TLR3–4 or the empty cloning vector pcDNA3.1+ showed TLR3 surface expression for the chimeric TLR3–4 construct and endosomal expression for human wild-type TLR3 (Fig. 4b–d). MD-2-CD14-HEK-293 cells transfected with the empty pcDNA3.1 cloning vector showed some intrinsic endosomal TLR3 expression (Fig. 4c). Similar results were also observed for untransfected HEK-293 cells (Fig. 2) and are most probably the result of endogenous TLR3 expression in those cells (Fig. 1). MD-2-CD14-HEK-293 cells transiently transfected with human TLR3 or chimeric TLR3–4 and activated with poly rI:rC or unformulated or formulated siRNA resulted in an immune response to poly rI:rC, but no activation when stimulated with unformulated or formulated siRNA (Fig. 4e). Longer or 2′O-methyl-modified siRNA also did not induce an immune response upon stimulation of human TLR3 or human TLR3–4 transfected MD2-CD14-HEK-293 cells (data not shown). Other reporter read-outs used, such as IFN-α, IFN-β, IL-8, IL-12, IP-10 (Interferon gamma-induced protein 10), Rantes (Regulated upon Activation, Normal T-cell Expressed, and Secreted) or tumour necrosis factor-α, could show activation by polyrI:rC in human TLR3 or TLR3–4 transfected MD2-CD14-HEK-293 cells, but failed to respond to unformulated or formulated siRNA (data not shown). In summary, we were unable to show a TLR3-dependent cellular response to unformulated or formulated siRNA via human surface TLR3 in vitro.

TLR3 inhibition by single-stranded DNA is CpG- and sequence-independent but backbone-dependent

Blockage of poly rI:rC-mediated TLR3 immune response by single-stranded DNA was further addressed in the second part of this study. Previous studies demonstrated inhibition of poly rI:rC-mediated TLR3 activation by single-stranded ODN.⁵¹ We were able to expand these studies by using the stable transfected HEK-293-NFkB-luc cells expressing TLR3. The cells were treated with poly rI:rC or a variety of different CpG and non-CpG-ODN and analysed for the concentration-dependent poly rI:rC inhibition. CpG-ODN were originally described as immune activating ligands for TLR9.^8,9 Two different parent ODN were used for the analysis: B-Class ODN 2006 (Table 1) and C-Class ODN 2395 (Table 2). We introduced sequence and backbone modifications in the CpG motif or the surrounding sequence of both ODN. CpG ODN 2006 contains four CpG motifs and CpG ODN 2395 contains two main CpG motifs and four additional ones in the palindromic tail.^8,52 Calculation of IC₅₀ values, as well as of the % inhibition of each altered ODN for their inhibitory effect on poly rI:rC, revealed for most ODN only minor differences without significant relevance (Tables 1 and 2; P-values of % inhibition). ODN 5276, for example, contained exchanges of all CpG ODN motif-surrounding thymidines by deoxy-uridines and showed a similar IC₅₀ (0·98 versus 0·71 μm) and % inhibition (77% versus 69%) to the parent sequence ODN 2006 (Table 1). These data imply that alterations to the surrounding sequence of CpG motifs do not show an effect on TLR3 inhibition. Exchanges of all four CpG motifs of ODN 2006 (as in ODN 2131) by GpC or only single nucleotide exchanges in the CpG motif through inosine (5249), 5-methyl-2′-deoxycytidine (2117), thymidine (5231) or even an abasic D-spacer (5280) did not reveal significant differences in single-stranded DNA-mediated TLR3 inhibition compared with the parent ODN 2006 (Table 1). Sequence alterations of both the CpG motif and the surrounding sequence as for ODN 5468, 20451 or 10105 also demonstrated no significant difference to ODN 2006 in terms of TLR3 inhibition. Similar data were observed for the parent C-Class CpG ODN 2395 (Table 2).

Table 1.

Toll-like receptor 3 (TLR3) inhibition by single-stranded DNA is backbone-dependent but sequence-independent.

Table 2.

Toll-like receptor (TLR3) inhibition by single-stranded DNA is sequence-independent but negatively influenced by the adenosine content.

Interestingly, exchange of nucleotides in the CpG motif or the surrounding sequence of ODN 2006 by the nucleotide adenosine led to loss of ODN-mediated TLR3 inhibition (ODN 5736, 20729, 5453; Tables 1 and 2) compared with the parent CpG ODN 2006 or 2395. 5736 revealed that the thymidine following the guanosine (e.g. in the CpG motif) cannot be replaced by adenosine. However, replacement of guanosine but not cytosine by adenosine is tolerated (compare ODN 5271 and 20730 with 20729 and 5453). We therefore assume that adenosine nucleotides at certain ODN positions can have a negative impact on ODN-mediated TLR3 inhibition.

Figure 5(a) shows that ODN containing only thymidine (ODN 5162) or adenosine (ODN 5163) are not able to block poly rI:rC-mediated TLR3 activation. Using AACGTT as a potential minimal sequence for TLR3 blockage in this context demonstrated again that adenosine reverses the inhibitory effect of ODN. Whereas sequence motifs for single-stranded DNA to inhibit TLR7, TLR8 and TLR9 activation were already described^34,53,54 we suggest that inhibiting a TLR3 immune response does not follow certain motif rules, but that adenosine at certain positions in a phosphorothioate ODN can reverse the inhibitory effect.

Figure 5.

Human Toll-like receptor 9 (TLR9) but not TLR3, TLR7 or TLR8 activation is regulated by CD14. HEK-293 cells stably transfected with human TLR3 (a) TLR7 (b), TLR8 (c) or TLR9 (d and e) were pre-incubated with or without sCD14 30 min before stimulation with the indicated concentrations of the listed TLR ligands. ORN R-0006 was additionally complexed to DOTAP (50 μg/ml→1/3 dilution). After 16 hr nuclear facotr-κB (NF-κB) activation was measured by assaying luciferase activity. Results are given as fold induction above background (medium). Sequences of CpG oligodeoxynucleotides (ODN) and oligoribonucleotides (ORN) are listed. Data are representative of five independent experiments. •, phosphorothioate backbone, -, phosphodiester backbone.

We further investigated the effect of the phosphorothioate backbone on TLR3 inhibition. ODN 5746, 5011 and 2059 mainly differ from the parent ODN 2006 in their backbone chemistry containing phosphodiester instead of phosphorothioate backbones (Table 1). Analysing the TLR3 inhibition induced by these ODN demonstrated a significant loss of inhibition for all three phosphodiester ODN. Phosphodiester ODN display lower stability and faster degradation. It is possible that either the greater instability is responsible for the lack of an inhibitory effect, or only phosphorothioate chemically modified, but not natural, linkages are capable of inhibiting TLR3 responses. To expand this observation, we started to analyse the effect of the length of a phosphorothioate ODN.

Our data demonstrate that sequence alterations by any nucleotide (other than adenosine at certain positions) do not have an impact to the potency of TLR3 blockage by single-stranded DNA. Therefore the parent ODN sequence 2006 was chosen to analyse potential length requirements. As shown in Fig. 5(b), 3′ deletions of ODN 2006 resulted in total loss of inhibitory activity starting with 12mers, whereas 14mers still resulted in decreased inhibition for poly rI:rC-mediated TLR3 activation. Similar data were obtained for 5′ deletions of ODN 2006 (data not shown).

In summary, we show that TLR3 blockage by single-stranded DNA is sequence-, length- and CpG-independent but backbone-dependent.

TLR9-dependent, but not TLR3-, TLR7- or TLR8-dependent, immune stimulation is regulated by CD14

The third part of this manuscript analyses the role of CD14 as potential co-receptor of the TLR3, TLR7, TLR8 or TLR9 immune response. Cells expressing TLR3, TLR7, TLR8 and TLR9 were stimulated by specific ligands and co-treated with CD14. CD14 is a glycosylphosphatidylinositol-anchored, membrane-associated protein that functions as an essential component of the TLR4/MD2 complex in LPS signalling.^27,28,55 It has also been described as an uptake enhancer for various TLR ligands including LPS, ceramide, poly rI:rC or double-stranded RNA.^31,56,57 In addition, CD14 has been proposed to mediate the uptake of poly rI:rC in murine cells expressing TLR3. Enhancement of the poly rI:rC, R-848 or B-Class CpG ODN activation in cells expressing murine TLR3 or TLR7, as well as murine and human TLR9, was also described and explained by direct binding of CD14 to these TLRs.^29,31 CD14 exists in two forms: anchored into the membrane (mCD14) or in soluble form (sCD14).

To investigate a potential effect of human CD14 on different TLRs, we first performed transient transfections of human mCD14 or sCD14 plasmids in stable transfected HEK-293-NFkB-luc cells expressing human TLR3, TLR7, TLR8 or TLR9. However, none of these TLR ligands in combination with CD14 showed enhanced immune stimulation compared with the same cells transfected with an empty cloning vector (data not shown). As another potential way to investigate the involvement of CD14 in human TLR activation ²⁹ we used human sCD14 protein to treat HEK-293-NFkB-luc cells expressing TLR3, TLR7, TLR8 or TLR9 (Fig. 5). Only TLR9-NFkB-HEK-293 cells in the presence of sCD14 showed strong enhancement of TLR9-mediated activation induced by CpG ODN 10103 (Fig. 5d). In contrast, no influence of sCD14 was observed for cells expressing human TLR3, TLR7 or TLR8 using specific agonists: poly rI:rC (TLR3), R-848 (TLR7 and TLR8), Sumitomo (TLR7) and/or ORN R-0006 (TLR7 and 8) (Fig. 5a–c). Co-culturing of cells for 8 or 48 hr with sCD14 and TLR ligand, or alteration of the pre-treatment time with sCD14 from 30 min up to 2 hr did not influence this result (data not shown).

Figure 6.

Toll-like receptor 3 (TLR3) blockage by single-stranded DNA is length-dependent and negatively influenced by adenosine nucleotides. HEK-293 cells stably transfected with human TLR3 were incubated with EC₅₀ concentrations of poly rI:rC (20 μg/ml) alone, or EC₅₀ concentrations of poly rI:rC or different oligodeoxynucleotides (ODN) (10 μm→1/3 dilution). Nuclear factor-κB (NF-κB) activation was measured by assaying luciferase activity. Results are given as fold induction above background (medium) (a) or % fold induction of EC₅₀ concentration of poly rI:rC (b). Mean of four independent experiments (n = 4). *Phosphorothioate backbone.

Human TLR9 can be activated by different CpG ODN classes, which differ in their cytokine and chemokine profiles and their structural properties and stabilities.⁵² CpG A-Class ODN contain poly G tails which have been shown to self-associate via Hoogsteen base-pairing to form parallel quadruplex structures called G-tetrads.^58,59 The formation of multimers is necessary for the localization of ODN in early endosomes and for their higher stability.^52,60–62 CpG C-Class ODN form perfect palindromes with stable dimers and were also described as being highly stable.⁵² CpG B-Class ODN, in contrast, are monomers lacking a secondary structure and with lower stability than A-Class or C-Class ODN (52 and data not shown). In addition, backbone modifications from phosphorothioate to phosphodiester in certain CpG ODN classes can alter their immune stimulatory potential and lower the ODN stability.^63–65 The TLR9 immune response of CpG ODN 10103 was strongly up-regulated by using sCD14 as co-receptor. This effect is clearly CpG-dependent because a control ODN (1982) without CpG ODN motif was ineffective in the presence of sCD14 (Fig. 5d). We further investigated sCD14-mediated enhancement of the human TLR9 immune response using different CpG ODN classes. Human TLR9-NFkB-HEK-293 cells were cultured in the presence or absence of sCD14 and stimulated with A-, B- or C-Class CpG ODN. An increase of the TLR9-mediated activation was detected for all CpG ODN classes (8954, 10103 and 2395) in the presence of sCD14. However, sCD14 treatment affected both the maximal activation and the potency of B-class ODN-mediated TLR9 stimulation, whereas stimulation by A- or C-Class ODN was only enhanced for the maximum activation, not the potency. B-Class ODN showed the strongest enhancement in TLR9 immune activation compared with A- and C-Class ODN. Interestingly, the presence of phosphodiester bonds in a B-Class ODN (5746) negatively influenced the capability of sCD14 to increase signalling, ODN-mediated TLR9 stimulation was rather down-regulated. In summary, our data suggest that sCD14 can differentially regulate CpG ODN-mediated human TLR9 activation.

Discussion

TLR7, TLR8 and TLR9 belong to a subfamily of TLRs and evolved as nearest neighbours. Specific recognition of nucleic acids such as single-stranded RNA for TLR7 and TLR8 or single-stranded CpG-ODN for TLR9 has already been described.^5–9,66 TLR3 evolved divergently from TLR7, TLR8 and TLR9, but also responds to nucleic acids like double-stranded RNA or its chemical mimic poly rI:rC.^15,16

TLR7, TLR8 and TLR9 are expressed in the intracellular endosomes,² whereas TLR3 has been described as being expressed in endosomes and on the cell surface.^10–14 Unformulated siRNA was recently postulated as a surface murine TLR3 ligand and as interfering with RNAi in a murine choroidal neovascularization model.^12,22 Therefore, we were interested to analyse several human cell lines for their TLR, RIG-I-like receptors, PKR and CD14 mRNA as well as for TLR3 endsomal or surface expression. We identified several cell lines with strong TLR3 surface expression including A549, APRE-19, HDMEC, HUVEC, JIMT1 and N87. We were also able to show that TLR3 surface expression is not limited to endothelial cell types and cell lines¹² but can be expanded to tumorigenic cell lines as well.

Selected cell lines with endosomal or surface TLR3 expression were incubated with poly rI:rC, or unformulated or formulated siRNA. Whereas the described TLR3 ligand poly rI:rC stimulated all TLR3 expressing cell lines, only cell lines A549, APRE-19 and RL-95-2 showed a response to formulated siRNA, which may be the result of TLR3. However, an influence of RIG-I, Mda-5 or PKR in the observed effects cannot be excluded because of the presence of the respective mRNAs. Therefore, we decided to over-express surface TLR3 using a chimeric TLR3–TLR4 construct in MD2-CD14-HEK-293 cells. Similar to the data above, poly rI:rC induced an immune activation via surface and endosomal TLR3, but we were unable to detect any TLR3 activation by unformulated or formulated siRNA. Previous results by Kleinman et al.¹² demonstrating activation of murine surface TLR3 by unformulated siRNA could not be reproduced for human TLR3. The reason may be differences in siRNA recognition between human and murine TLR3. Binding of double-stranded RNA to the TLR3 ectodomain occurs at amino acids H539 and N541 in leucine-rich repeat (LRR) 20. Loop1 (residues 335–343) in LRR12 of the ectodomain is responsible for TLR3 activity.^10,13,51,67 As described in the literature, loop 1 is variable in length and sequence composition among species ¹³ and murine TLR3 contains two amino acid exchanges to QSVSLASH instead of QSISLASL. Poly rI:rC immune activation via murine and human TLR3 appears similar,¹³ but we speculate that the differences between human (our data) and murine ^12,22,23 TLR3 in activation induced by unformulated siRNA is the result of these sequence differences.

CD14 is a well-known pattern recognition receptor in the innate immune system and is described as a co-receptor via ligand binding for TLR4, TLR3, TLR7 and TLR9.^27,29,31 Beside TLR9 and TLR4, all other published results were generated using murine TLRs. Therefore, we used human TLR3, TLR7, TLR8 and TLR9, and could not confirm an enhancement of TLR3, TLR7 or TLR8 immune activation using human CD14 as a potential co-receptor. This may be because of differences in CD14 regulation between different species. The TLR9 up-regulation through CD14 is clearly CpG ligand-dependent because no effect using control ODN was observed.

Investigation of the enhancement of human TLR9 activation by CD14 was expanded to include several CpG ODN classes: the A-, B- and C-Classes. These classes differ in their structural properties and immune stimulatory profile: G-tails in A-Class ODN have been shown to self-associate via Hoogsteen base-pairing to form parallel quadruplex structures called G-tetrads.^58,59 B-Class ODN lack secondary structures and C-Class ODN form perfect palindromes with stable dimers.⁵² Interestingly, TLR9 activation was differentially regulated by sCD14: B-Class, followed by A-Class and C-Class CpG ODN showed an enhanced effect in the presence sCD14. We speculate, therefore, that because of tertiary or secondary structure in A-Class and C-Class ODN, interaction between the CD14 protein, the TLR9 receptor and its ligand is not as favourable as for linear B-Class.

Beside these classical B-Class and C-Class ODN we used phosphodiester versions of both classes. Phosphodiester ODN in combination with CD14 did not result in an enhanced immune response. Here, a decrease of TLR9-mediated effects could be observed. Phosphodiester linkages result, especially in ODN without stabilizing secondary or tertiary structures, in decreased stability and lower protection against endonucleases or exonucleases.^52,68,69 Therefore, lower stability of phosphodiester linkages in CpG ODN without secondary or tertiary structure may be a reason for the lack of an observable effect on human TLR9.

Blockage of TLR7, TLR8 and TLR9 by single-stranded DNA containing certain suppressive motifs has been described previously.^34,54,70 Recently, it has been published that immune activation via TLR3 can also be blocked by single-stranded DNA.⁵¹ However, no specific sequence requirements or inhibitory motif or a more detailed analysis for this TLR3 inhibition were described. We demonstrate that poly rI:rC-mediated TLR3 activation indeed can be blocked by single-stranded DNA, but in a backbone-dependent and length-dependent manner. Single-stranded DNA with 12 or fewer nucleotides or with phosphodiester linkages failed to block TLR3 immune effects. Alteration of the inhibitory single-stranded DNA by nucleotides, abasic spacer or its chemical modifications do not influence TLR3 blockage. This implies that in contrast to the blockage of TLR7, TLR8 and TLR9 by specific sequence motifs of single-stranded DNA,^34,53,54,70 TLR3 does not follow the same motif rules. However, adenosine insertions at certain positions of single-stranded DNA can result in loss of the inhibitory effect. Adenosines can be anti-inflammatory through the A(2A) receptor.^71–74 Synergistic interaction has been reported ^75,76 between the A(2A) receptor and the receptors TLR2, TLR3, TLR4, TLR7 and TLR9, resulting in down-regulation of tumour necrosis factor-α release upon murine macrophage activation and human peripheral blood mononuclear cell stimulation. It appears therefore possible that a high adenosine content or the presence of adenosines at certain positions abolishes TLR3 blockage through possible binding of the ODN to the A(2A) receptor. Combination of TLR3 with already published sequence information and motif rules for TLR7, TLR8 and TLR9 may have great impact on clinical development of single-stranded DNA to treat autoimmune diseases like systemic lupus erythematosus or rheumatoid arthritis.

In summary, we show that it is not possible to observe an immune response to unformulated siRNA by surface human TLR3 as suggested before for murine TLR3. This may be because of differences in TLR3 protein loop 1 between species, which is responsible for TLR3 activation. The putative co-receptor CD14 appears to function as a signalling regulator only for human TLR9, but not for human TLR3, TLR7 or TLR8. In addition, TLR3 activation can be blocked by single-stranded DNA in a backbone- and length-dependent but sequence- (except for Adenosines at certain positions) and CpG-independent manner.

Disclosure

There are no concerns for public disclosure.