Abstract
Human TLR7 and 8 (hTLR7/8) have been implicated in the sequence-dependent detection of RNA oligonucleotides in immune cells. Although hTLR7 sequence-specific sensing of short RNAs has been inferred from studies of murine TLR7, this has yet to be established for hTLR7. We found that different short ssRNA sequences selectively induced either TNF-α or IFN-α in human PBMCs. The sequence-specific TNF-α response to ssRNAs observed in PBMCs could be replicated in activated human macrophage-like (THP-1) cells pretreated with IFN-γ. Surprisingly, suppression of hTLR7 expression by RNA interference in this model reduced sensing of all immunostimulatory ssRNAs tested. Modulation of the relative expression ratio of hTLR7 to hTLR8 in THP-1 cells correlated with differential sensing of immunostimulatory sequences. Furthermore, the sequence-specific IFN-α induction profile in human PBMCs was accurately modeled by a sequence-specific activation of murine TLR7 in mouse macrophages. Thus, we demonstrate for the first time that hTLR7 is involved in sequence-specific sensing of ssRNAs. We establish a novel cell model for the prediction of TNF-α induction by short RNAs in human macrophages. Our results suggest that differential sequence-specific sensing of RNA oligonucleotides between human and mouse macrophages is due to the modulation of TLR7 sensing by human TLR8.
Toll-like receptors are a family of proteins recognizing different pathogen-associated molecular patterns and form part of the first line of defense against pathogens. Once engaged by the TLRs, pathogen-associated molecular patterns trigger a signaling cascade resulting in specific gene expression profiles that stimulate the immune system to eliminate the invading pathogen (1). Among the 10 TLRs reported in humans, TLR3, 7, 8, and 9 constitute a subgroup that can detect bacterial or viral nucleic acids.
Human TLR7 (hTLR7)3 and TLR8 (hTLR8) are located in the endosomes, are closely related in structure, and have been directly implicated in the detection of RNA by genetic complementation in HEK293 cells (2, 3). Among PBMCs, it is believed that hTLR7 is expressed predominantly in plasmacytoid dendritic cells (pDCs) and B cells, while hTLR8 is mainly expressed in monocytes and myeloid dendritic cells (4, 5). However, microarray data mining shows that hTLR7 is also expressed in monocytes, albeit at much lower levels relative to hTLR8 (6). Similarly, hTLR8 expression occurs in pDCs.
Studies characterizing hTLR7 and 8 have typically used synthetic immune response modifiers as ligands, such as imidazoquinolines (7, 8), and show that hTLR7 agonists mainly stimulate pDCs to produce IFN-α, while hTLR8 agonists activate monocytes to produce proinflammatory cytokines, notably TNF-α (9, 10). Consistent with this, PBMCs depleted of pDCs do not produce IFN-α upon RNA stimulation (11).
To date, the role of hTLR7 in RNA sensing remains poorly defined. Studies using an overexpression system in HEK293 cells failed to implicate hTLR7 in sequence-dependent sensing of ssRNA oligonucleotides (11, 12). In contrast, similar experiments demonstrated that hTLR8 was involved in uridine-rich ssRNA sensing (11, 12). Because the murine homolog of TLR7 (mTLR7) is required for sequence-specific ssRNA sensing, Heil et al. (11) concluded that species-specific differences exist in the sensing of RNA oligonucleotides where hTLR8 but not hTLR7 is the sequence-specific receptor of ssRNAs in humans. Recent studies investigating the immunostimulation of small interfering RNA (siRNA) in immune cells have suggested a role for hTLR7 in sequence-dependent sensing of siRNA and its ssRNA components on the basis of IFN-α production by immune cells. However, no mechanistic evidence was provided, showing that hTLR7 was involved in the responses observed (13, 14, 15, 16, 17, 18).
In the present study, we use a human macrophage-like cell line and provide direct evidence that hTLR7 is involved in the sequence-dependent sensing of ssRNAs.
Materials and Methods
Cell isolation and culture
Fresh blood from healthy male donors was collected in heparin-treated tubes, and submitted to Ficoll-Paque plus (17-1440-02; GE Healthcare) gradient purification following the manufacturer’s guidelines. Isolated cells were plated in a 96-well plate at 2 × 105 cells/well in RPMI 1640 plus l-glutamine medium (11875; Invitrogen Life Technologies) complemented with 1× antibiotic/antimycotic (15204064; Invitrogen Life Technologies) and 10% FBS (referred to as complete RPMI 1640), and incubated for 4 h at 37°C in a 5% CO2 atmosphere before stimulation with TLR agonists. Human THP-1 and U937 cells were maintained in complete RPMI 1640 and subcultured in suspension every 2–3 days until passage (p < 25). For experiments, 80,000 THP-1 cells were differentiated for 16 h in conditioned medium with PMA at 20 ng/ml (in DMSO 524400; Calbiochem) per well of a 96-well plate (at 37°C in a 5% CO2 atmosphere). Priming of adherent PMA-treated THP-1 was conducted by rinsing the cells with 150 μl of complete RPMI 1640 supplemented with 100 U/ml human IFN-γ (IF002, ED50 < 0.05 ng/ml; Chemicon International) or 50 ng/ml recombinant human IL-6 (a gift from B. J. Jenkins, Monash Institute of Medical Research, Clayton, Victoria, Australia), for 6 h at 37°C, 5% CO2. The cells were subsequently rinsed with 100 or 150 μl of complete RPMI 1640 before stimulation with TLR agonists. The RAW-ELAM cells, stably transfected with an ELAM promoter driving the expression of luciferase (19), were cultured in DMEM complemented with 10% FCS and 1× antibiotic/antimycotic. RAW-ELAM cells were plated to 90% confluency in a 96-well plate in 150 μl of culture medium for 7 h, and further stimulated with TLR agonists. The supernatants were collected for TNF-α analysis 14 h posttransfection.
Isolation of bone marrow macrophages
TLR7-deficient mice (crossed to the BALB/c background for three generations (N3), a gift from Dr. S. Akira, Research Institute for Microbial Diseases, Osaka University, Osaka, Japan) were crossed to the BALB/c background for a further four generations (N7). Wild-type BALB/c mice generated from the original stock of TLR7-deficient mice were treated identically and used as controls. Mice were housed at the University of Newcastle and experiments approved under ACEC939. Bone marrow extraction and differentiation was conducted following standard procedures (20). Briefly, femurs were flushed with complete RPMI 1640, and cells were plated in complete RPMI 1640 supplemented with 104 U/ml recombinant human CSF-1 (a gift from D. A. Hume, Centre for Molecular Biology and University of Biotechnology, Queensland, Brisbane, Australia) on 10-cm bacteriological plastic plates for 7 days at 37°C in a 5% CO2 atmosphere.
Cell stimulation
ssRNAs were synthesized by Integrated DNA Technologies (IDT) with HPLC purification, and resuspended into duplex buffer (100 mM potassium acetate, 30 mM HEPES (pH 7.5), DNase-RNase-free H2O) to a concentration of 80 μM. ssRNAs from different syntheses were compared and no immunostimulatory difference was observed. 3M-002 (TLR8 agonist; tlrl-c75), loxoribine (TLR7 agonist), and ODN2216 (TLR9 agonist; tlrl-hodna) were purchased from Invivogen and LPS was obtained from Sigma- Aldrich (L8274). ODN2216, loxoribine, 3M-002, and LPS were added directly to medium at the specified concentration. ssRNAs were transfected into cells with N-[1-(2,3-Dioleoyloxy)propyl]-N,N,Ntrimethylammonium methylsulfate (DOTAP) (1811177; Roche). DOTAP was first diluted in RPMI 1640 (75 μl) for 5 min before mixing with an equal volume of RPMI 1640 containing the ssRNA. The resulting mix was incubated >10 min and 50 μl were added per well of a 96-well plate, resulting in a final volume of 150 or 200 μl. Transfections were conducted in triplicate in all experiments. The ratios of DOTAP to ssRNA were as follows: 5.3 μg/μl ssRNA (see Figs. 1 and 5,b), 2.8 μg/μl ssRNA (see Fig. 2), and 3.74 μg/μl ssRNA (see Figs. 3, 4, 5,c, and 6). When dose responses were conducted, the ratio of ssRNA to DOTAP was maintained constant.
Sequence-dependent differential induction of IFN-α and TNF-α by ssRNA in human PBMCs. a and c, Human PBMCs were treated with 90 nM of indicated ssRNA complexed with DOTAP and incubated for 16 h at 37°C. Cytokine production (IFN-α and TNF-α) was measured by ELISA as described in Materials and Methods. 3M-002 (1 μg/ml) and ODN2216 (3 μM) were used as positive controls for induction of TNF-α and IFN-α, respectively. b, Schematic of base substitutions of ssRNA B-406-S. The secondary structure was conserved between the ssRNAs, as predicted with mFOLD (24 ). The data (a and c) are in biological triplicate and are representative of a minimum of three different blood donors.
Sequence-dependent differential induction of IFN-α and TNF-α by ssRNA in human PBMCs. a and c, Human PBMCs were treated with 90 nM of indicated ssRNA complexed with DOTAP and incubated for 16 h at 37°C. Cytokine production (IFN-α and TNF-α) was measured by ELISA as described in Materials and Methods. 3M-002 (1 μg/ml) and ODN2216 (3 μM) were used as positive controls for induction of TNF-α and IFN-α, respectively. b, Schematic of base substitutions of ssRNA B-406-S. The secondary structure was conserved between the ssRNAs, as predicted with mFOLD (24 ). The data (a and c) are in biological triplicate and are representative of a minimum of three different blood donors.
TLR7/8 induction by IFN-γ in monocytic cells correlates with sequence-specific sensing of ssRNAs. a, CD14+ monocytes and U937 cells were treated with 100 U/ml IFN-γ for 6 and 8 h, respectively, and RNA was extracted for real-time RT-PCR analysis. TLR7 and TLR8 mRNA levels are shown, reported relative to GAPDH expression, and further divided by the mean level of nontreated cells. The data for CD14+ cells are from one blood donor in biological triplicate and are representative of two blood donors. The data for U937 cells are averaged from three independent experiments in triplicate. b, CD14+ monocytes were purified as presented in Materials and Methods, plated in complete RPMI 1640 at 50,000–100,000 cells/well of a 96-well plate and primed (+IFN-γ) or not (Medium) for 6 h with 100 U/ml IFN-γ. After rinsing with fresh complete RPMI 1640 medium, the cells were incubated with the indicated ssRNA (90 nM) complexed with DOTAP. Sixteen hours after treatment, the levels of TNF-α and IFN-α (data not shown) were analyzed by ELISA. The data are averaged from two independent experiments in triplicate, from two different blood donors. Statistical analyses comparing conditions si9.2-S and SA are presented (NS, nonsignificant). c, U937 cells were plated in growing medium complemented with 100 U/ml IFN-γ (+IFN-γ) or not (+Medium) at 80,000 cells/well of a 96-well plate. After an 8-h incubation, the indicated ssRNAs complexed with DOTAP were added and incubated for 16 h before analysis of TNF-α levels via ELISA. The data are representative of two independent experiments in triplicate.
TLR7/8 induction by IFN-γ in monocytic cells correlates with sequence-specific sensing of ssRNAs. a, CD14+ monocytes and U937 cells were treated with 100 U/ml IFN-γ for 6 and 8 h, respectively, and RNA was extracted for real-time RT-PCR analysis. TLR7 and TLR8 mRNA levels are shown, reported relative to GAPDH expression, and further divided by the mean level of nontreated cells. The data for CD14+ cells are from one blood donor in biological triplicate and are representative of two blood donors. The data for U937 cells are averaged from three independent experiments in triplicate. b, CD14+ monocytes were purified as presented in Materials and Methods, plated in complete RPMI 1640 at 50,000–100,000 cells/well of a 96-well plate and primed (+IFN-γ) or not (Medium) for 6 h with 100 U/ml IFN-γ. After rinsing with fresh complete RPMI 1640 medium, the cells were incubated with the indicated ssRNA (90 nM) complexed with DOTAP. Sixteen hours after treatment, the levels of TNF-α and IFN-α (data not shown) were analyzed by ELISA. The data are averaged from two independent experiments in triplicate, from two different blood donors. Statistical analyses comparing conditions si9.2-S and SA are presented (NS, nonsignificant). c, U937 cells were plated in growing medium complemented with 100 U/ml IFN-γ (+IFN-γ) or not (+Medium) at 80,000 cells/well of a 96-well plate. After an 8-h incubation, the indicated ssRNAs complexed with DOTAP were added and incubated for 16 h before analysis of TNF-α levels via ELISA. The data are representative of two independent experiments in triplicate.
Cytokine priming of THP-1 cells restores functional responsiveness to ssRNAs. a, PMA-differentiated THP-1 cells were treated for 8 h with the indicated amount of IFN-γ and RNA extracted for real-time RT-PCR analysis. TLR7 and TLR8 mRNA levels were reported relative to GAPDH expression, and further divided by the mean level of the PMA-differentiated condition. Each data point is from biological triplicates, and the data are representative of two independent experiments. b, PMA-differentiated THP-1 pretreated with IFN-γ were stimulated with 330 nM of indicated ssRNA complexed with DOTAP and incubated for 16 h. 3M-002 (2 μg/ml) and loxoribine (500 μM) were included as controls for TNF-α production as measured by ELISA. The data are averaged from a minimum of two independent experiments, where each treatment was conducted in triplicate and reported to the mean of B-406-AS condition. All ssRNA significantly induced TNF-α production when compared with the mock control (p < 0.001) unless otherwise noted (NS).
Cytokine priming of THP-1 cells restores functional responsiveness to ssRNAs. a, PMA-differentiated THP-1 cells were treated for 8 h with the indicated amount of IFN-γ and RNA extracted for real-time RT-PCR analysis. TLR7 and TLR8 mRNA levels were reported relative to GAPDH expression, and further divided by the mean level of the PMA-differentiated condition. Each data point is from biological triplicates, and the data are representative of two independent experiments. b, PMA-differentiated THP-1 pretreated with IFN-γ were stimulated with 330 nM of indicated ssRNA complexed with DOTAP and incubated for 16 h. 3M-002 (2 μg/ml) and loxoribine (500 μM) were included as controls for TNF-α production as measured by ELISA. The data are averaged from a minimum of two independent experiments, where each treatment was conducted in triplicate and reported to the mean of B-406-AS condition. All ssRNA significantly induced TNF-α production when compared with the mock control (p < 0.001) unless otherwise noted (NS).
ssRNAs induce TNF-α release through TLR7- and TLR8-dependent pathways in THP-1 cells. PMA-differentiated THP-1 cells were transfected with the indicated siRNAs (17 nM), before further treatment with IFN-γ and ssRNAs as detailed in Materials and Methods. a, Real-time RT-PCR analysis of TLR7 and TLR8 expression following 8 h of IFN-γ priming, reported relative to GAPDH levels and expressed as a percent of expression of siControl treatment. The Neg condition indicates the levels of TLR7 and TLR8 independent of IFN-γ priming. The data are from biological triplicates and are representative of two independent experiments. b and c, TNF-α levels were measured after treatment with TLR7 (loxoribine at 900 μM), TLR8 (3M-002 at 1 μg/ml), and TLR4 (LPS at 500 ng/ml) agonists (b) or ssRNAs (at 660 nM) in the presence of siRNAs (c). The data are averaged from two (c) or three independent experiments in triplicate (b). Statistical analysis is as presented using siGAPDH treatment as a reference.
ssRNAs induce TNF-α release through TLR7- and TLR8-dependent pathways in THP-1 cells. PMA-differentiated THP-1 cells were transfected with the indicated siRNAs (17 nM), before further treatment with IFN-γ and ssRNAs as detailed in Materials and Methods. a, Real-time RT-PCR analysis of TLR7 and TLR8 expression following 8 h of IFN-γ priming, reported relative to GAPDH levels and expressed as a percent of expression of siControl treatment. The Neg condition indicates the levels of TLR7 and TLR8 independent of IFN-γ priming. The data are from biological triplicates and are representative of two independent experiments. b and c, TNF-α levels were measured after treatment with TLR7 (loxoribine at 900 μM), TLR8 (3M-002 at 1 μg/ml), and TLR4 (LPS at 500 ng/ml) agonists (b) or ssRNAs (at 660 nM) in the presence of siRNAs (c). The data are averaged from two (c) or three independent experiments in triplicate (b). Statistical analysis is as presented using siGAPDH treatment as a reference.
Differential sensing of ssRNAs correlates with differential expression ratio of hTLR7 to hTLR8. a, PMA-differentiated THP-1 cells were treated for 8 h with the indicated amount of IL-6, and RNA was extracted for real-time RT-PCR analysis. TLR7 and TLR8 mRNA levels are shown reported relative to GAPDH expression, and further divided by the mean level of the PMA-differentiated condition. Each data point is from biological triplicates, and the data are representative of two independent experiments. b and c, PMA-differentiated THP-1 were primed with 100 U/ml IFN-γ (b) or 50 ng/ml IL-6 (c) for 8 h before further stimulation with the indicated ssRNA/DOTAP complexes. Sixteen hours after treatment, the levels of TNF-α were analyzed by ELISA. The data are averaged from two independent experiments in triplicate and are representative of four independent experiments. Statistical analysis comparing SA and si9.2-S is presented. d and e, Human PBMCs were subjected to a similar dose-response treatment of ssRNAs as in b and c, and the supernatants were analyzed for IFN-α and TNF-α after 16 h of treatment. Each treatment was conducted in triplicate, and the data presented are from two different blood donors. No IFN-α was detectable (ND) with SA and B-406-S (e). 3M-002 (1 μg/ml) (b–e) and loxoribine (900 μM) (b and c) were used as positive controls for TLR7/8 activation. Mock control corresponds to the highest amount of DOTAP used with 1 μM ssRNA. The dose responses were conducted with a constant ratio of DOTAP/ssRNA.
Differential sensing of ssRNAs correlates with differential expression ratio of hTLR7 to hTLR8. a, PMA-differentiated THP-1 cells were treated for 8 h with the indicated amount of IL-6, and RNA was extracted for real-time RT-PCR analysis. TLR7 and TLR8 mRNA levels are shown reported relative to GAPDH expression, and further divided by the mean level of the PMA-differentiated condition. Each data point is from biological triplicates, and the data are representative of two independent experiments. b and c, PMA-differentiated THP-1 were primed with 100 U/ml IFN-γ (b) or 50 ng/ml IL-6 (c) for 8 h before further stimulation with the indicated ssRNA/DOTAP complexes. Sixteen hours after treatment, the levels of TNF-α were analyzed by ELISA. The data are averaged from two independent experiments in triplicate and are representative of four independent experiments. Statistical analysis comparing SA and si9.2-S is presented. d and e, Human PBMCs were subjected to a similar dose-response treatment of ssRNAs as in b and c, and the supernatants were analyzed for IFN-α and TNF-α after 16 h of treatment. Each treatment was conducted in triplicate, and the data presented are from two different blood donors. No IFN-α was detectable (ND) with SA and B-406-S (e). 3M-002 (1 μg/ml) (b–e) and loxoribine (900 μM) (b and c) were used as positive controls for TLR7/8 activation. Mock control corresponds to the highest amount of DOTAP used with 1 μM ssRNA. The dose responses were conducted with a constant ratio of DOTAP/ssRNA.
ssRNAs induce TNF-α in a TLR7-dependent manner in mouse macrophages. a, RAW-ELAM cells were treated with the indicated doses of ssRNA/DOTAP complexes and incubated for 14 h before TNF-α analysis of the medium via ELISA. Mock control corresponds to the highest amount of DOTAP used with 750 nM ssRNA. The data are representative of three independent experiments in triplicate. b, BMMs from TLR7−/− and TLR7+/+ littermates were treated for 16 h with the indicated ssRNA/DOTAP doses (80 or 500 nM) and TNF-α was measured via ELISA. The Mock control for both 80 and 500 nM ssRNA is shown. 3M-002 (1 μg/ml) and LPS (500 ng/ml) were used as controls, and were not complexed with DOTAP (a and b). Statistical analyses shown are compared with Mock 500 nM (NS: nonsignificant). The data are averaged from three TLR7+/+ and three TLR7−/− mice, in triplicate, conducted in two independent experiments.
ssRNAs induce TNF-α in a TLR7-dependent manner in mouse macrophages. a, RAW-ELAM cells were treated with the indicated doses of ssRNA/DOTAP complexes and incubated for 14 h before TNF-α analysis of the medium via ELISA. Mock control corresponds to the highest amount of DOTAP used with 750 nM ssRNA. The data are representative of three independent experiments in triplicate. b, BMMs from TLR7−/− and TLR7+/+ littermates were treated for 16 h with the indicated ssRNA/DOTAP doses (80 or 500 nM) and TNF-α was measured via ELISA. The Mock control for both 80 and 500 nM ssRNA is shown. 3M-002 (1 μg/ml) and LPS (500 ng/ml) were used as controls, and were not complexed with DOTAP (a and b). Statistical analyses shown are compared with Mock 500 nM (NS: nonsignificant). The data are averaged from three TLR7+/+ and three TLR7−/− mice, in triplicate, conducted in two independent experiments.
RNA interference
siTLR8 was purchased from Dharmacon as an ON-TARGETplus SMARTpool (l-004715-00), and both siGAPDH (4631) and siControl 1 (4635) were purchased from Ambion. Duplexed siRNAs were resuspended following the manufacturers’ guidelines, and the sequences were not supplied. siTLR7 sequences previously published (21) are as follows: TLR7-sense (S) 5′-GCCUUGAGGCCAACAACAUdTdT-3′, TLR7-antisense (AS) 5′-AUGUUGUUGGCCUCAAGGCdTdT-3′ and were synthesized by IDT, resuspended into duplex buffer, annealed at 92°C for 2 min and left for 30 min at room temperature before aliquoting. All siRNAs were resuspended to a final concentration of 40 μM. A total of 80,000 THP-1 cells were differentiated for 10 h with PMA at 20 ng/ml per well of a 96-well plate. Subsequently, the cells were rinsed with 100 μl of RPMI 1640 plus l-glutamine medium (11875; Invitrogen Life Technologies) complemented with 10% FBS without antibiotics. Lipofectamine RNAiMAX (13778-075; Invitrogen Life Technologies) was first diluted in Opti-MEM (51985-034; 75 μl; Invitrogen Life Technologies) for 5 min before mixing with an equal volume of Opti-MEM containing the siRNA. After 20 min of incubation, 50 μl of the resulting RNAiMAX/siRNA was added directly onto the cells, giving a final volume of 150 μl. The ratio of Lipofectamine RNAiMAX to siRNA was 5-1 μl. Following a 14-h incubation at 37°C (5% CO2 atmosphere), the cells were further rinsed with 150 μl of complete RPMI 1640 supplemented with 100 U/ml IFN-γ. No cell death was observed during this procedure and the cells were further treated as described above.
Real-time RT-PCR
cDNA was synthesized from column-purified RNA (NucleoSpin RNAII columns; 740955; Macherey-Nagel) using the SuperScript III First-Strand kit (18080-051; Invitrogen Life Technologies), with oligo-dT(20) priming and following the manufacturer’s instructions. Real-time PCR was conducted with the iQ SYBR Green Supermix (170-8882; Bio-Rad) on a Bio-Rad iCycler. hTLR7 (NM_016562) and hTLR8 (NM_138636) were amplified using the following primer pairs (5′–3′): TLR7 forward (FWD): CCTTTCCCAGAGCATACAGC; TLR7 reverse (REV): GGACAGAACTCCCACAGAGC; TLR8-FWD: CAGAGCATCAACCAAAGCAA; TLR8-REV: GCTGCCGTAGCCTCAAATAC. hGAPDH (NM_002046) was used as a reference gene and amplified with the following primer pair: GAPDH-FWD: CATCTTCCAGGAGCGAGATCCC; GAPDH-REV: TTCACACCCATGACGAACAT. Each amplicon was sequence-verified and used to generate a standard curve for the quantification of gene expression.
Detection of cytokines
Human IFN-α in culture supernatants was quantified by sandwich ELISA using mouse monoclonal (0.5 μg/ml, 21112-1; PBL Biomedical) and rabbit polyclonal Abs (0.5 μg/ml, 31130-1; PBL Biomedical). A goat anti-rabbit HRP-conjugated Ab (0.8 μg/ml, 31460; Pierce) was used for detection. Human and mouse TNF-α were measured using the OptEIA ELISA sets (555212 and 558874, respectively; BD Biosciences). In both IFN and TNF-α ELISAs, tetramethyl benzidine substrate (T0440; Sigma-Aldrich) was used for quantification of the cytokines on the Fluostar OPTIMA (BMG Labtech) plate reader.
Magnetic purification of CD14+ monocytes
After extraction of PBMCs by Ficoll-Paque, cells were pelleted at 350 × g for 7 min and resuspended in magnetic buffer (2 mM EDTA, 0.5% BSA in PBS) at a final 108 cells/ml buffer. The cells were incubated with 100 μl of StemSep Human CD14 Positive Selection Cocktail (14758; StemCell Technologies) per milliliter of cell suspension for 10 min at 4°C. A total of 60 μl of magnetic colloid were subsequently added per milliliter of cell suspension and incubated for another 10 min at 4°C. The cell suspension was rinsed with 16 ml of magnetic buffer per milliliter of cell suspension and pelleted at 300 × g for 8 min. The pellet was resuspended in 4 ml of magnetic buffer per milliliter of cell suspension, before loading onto an LS column (130-042-401; Miltenyi Biotec) according to the manufacturer’s recommendations. The purity of eluted cells from positive magnetic purification was assessed by flow cytometer analysis using BD Pharmingen CD14 PE-conjugated Ab (BD Biosciences), and was >97%.
Statistical analyses
Statistical analyses were conducted using GraphPad Instat version 3.05. Unpaired t tests with Welch correction were used to compare the significance of the results. Error bars on each figure represent the SEM. Symbols used: ∗, p < 0.05; ∗∗, p < 0.01; ∗∗∗, p < 0.0001.
Results
Sequence-dependent differential induction of IFN-α and TNF-α by ssRNA
Different ssRNA sequences were screened for immunostimulatory activity in human PBMCs (6 of 14 sequences tested are shown in Fig. 1,a with the detail of the sequences in Table I), and it was observed that some sequences strongly induced both IFN-α and TNF-α (such as B-406-AS, Lamin-AS, and STAT-2AS), while others preferentially induced IFN-α (see si9.2-S and GFP21-AS). Because it has been previously suggested that IFN-α induction may be more specific to hTLR7 and that TNF-α may reflect more of an hTLR8 response (9), we speculated that preferential hTLR7 or hTLR8 signaling by RNA ligands may be sequence specific.
ssRNA sequences used in the study
Name of Sequence . | Sequence . | Uridines . |
---|---|---|
B-406-S (reported as βgal-478-S (13)) | 5′-GAAGGCCAGACGCGAAUUAUU-3′ | 4 |
B-406-AS (reported as βgal-478-AS (13)) | 5′-UAAUUCGCGUCUGGCCUUCUU-3′ | 9 |
Lamin-AS | 5′-UGUUCUUCUGGAAGUCCAGdTdT-3′ | 7 |
si9.2-S (reported as siRNA9.2-S (18)) | 5′-AGCUUAACCUGUCCUUCAAdTdT-3′ | 6 |
STAT-2AS | 5′-GUUCCAUUGGCUCUGGUGCUU-3′ | 9 |
GFP21-AS | 5′-GAUGAACUUCAGGGUCAGCUU-3′ | 6 |
SA | 5′-GAAGGCCUUACGCGAAUUAUU-3′ | 6 |
SB | 5′-GAAGGCCUUACGCGAACAAUU-3′ | 4 |
SC | 5′-GAAGGCCUUACGCGAACAAGA-3′ | 2 |
Name of Sequence . | Sequence . | Uridines . |
---|---|---|
B-406-S (reported as βgal-478-S (13)) | 5′-GAAGGCCAGACGCGAAUUAUU-3′ | 4 |
B-406-AS (reported as βgal-478-AS (13)) | 5′-UAAUUCGCGUCUGGCCUUCUU-3′ | 9 |
Lamin-AS | 5′-UGUUCUUCUGGAAGUCCAGdTdT-3′ | 7 |
si9.2-S (reported as siRNA9.2-S (18)) | 5′-AGCUUAACCUGUCCUUCAAdTdT-3′ | 6 |
STAT-2AS | 5′-GUUCCAUUGGCUCUGGUGCUU-3′ | 9 |
GFP21-AS | 5′-GAUGAACUUCAGGGUCAGCUU-3′ | 6 |
SA | 5′-GAAGGCCUUACGCGAAUUAUU-3′ | 6 |
SB | 5′-GAAGGCCUUACGCGAACAAUU-3′ | 4 |
SC | 5′-GAAGGCCUUACGCGAACAAGA-3′ | 2 |
To further characterize the differential activation of IFN-α and TNF-α observed with different ssRNAs, the impact of base substitutions on a sequence that was not immunostimulatory (B-406-S, Fig. 1,a) was assessed. It has been previously proposed that hTLR7/8 sensing of ssRNAs is dependent on the presence of uridine ribonucleotides, as observed in mice with mTLR7 (11, 16, 22, 23). Accordingly, we determined whether the degree of immunostimulation was simply a reflection of the number of uridines present, or whether the location of the uridines in the ssRNA relative to its secondary structure had any influence on immunostimulation. To do this, three ssRNAs were designed where the number of uridines was varied and their location relative to a central stem loop changed, without disrupting the predicted secondary structure (Fig. 1,b) (24). As expected, a significant induction of TNF-α was observed with SA (p < 0.0001, compared with B-406-S), which contains two more uridine residues than B-406-S, in the stem loop (Fig. 1, b and c). In contrast, however, SB induced more TNF-α than B-406-S (p = 0.0217), even though both have the same number of uridines. This is in accord with immunostimulation being affected by the location of the uridines within the ssRNA.
Unexpectedly, neither SA nor SB induced IFN-α production (Fig. 1,c). This contrasts with Lamin-AS, which, although it has a similar amount of uridine residues to SA (Table I), stimulates the release of both TNF-α and IFN-α (Fig. 1, a and c). Taken together, these results suggested that sequence-specific sensing of ssRNAs in PBMCs is mediated differentially by hTLR8 and hTLR7.
A human macrophage cell model reproduces functional sensing of ssRNAs
Because human macrophage-like THP-1 (25) cells naturally express hTLR7 and 8 and have been successfully used for studying synthetic hTLR7/8 agonists (26, 27, 28), they provide a natural model to study endogenous TLR signaling rather than the overexpression of the receptors in HEK293 cells. We found that ssRNA stimulation of THP-1 cells with or without PMA differentiation did not promote detectable induction of TNF-α on ELISA, although PMA treatment strongly induced sensitivity of the cells to the TLR4 ligand LPS (data not shown). Given the functionality of the TLR4/MyD88-signaling pathway in PMA-differentiated THP-1 cells, we reasoned that the absence of ssRNA sensing was related to low levels of hTLR7/8. In accord with this, PMA-differentiated THP-1 cells stimulated with synthetic hTLR7 agonists failed to induce TNF-α production (data not shown; Ref. 27), suggesting TLR7 was not expressed at a functional level in these cells. IFN-γ priming of PMA-differentiated THP-1 cells has been proposed to selectively induce both hTLR7 and hTLR8, with a predominance for hTLR8 (26). Accordingly, IFN-γ priming (at 50–100 U/ml) of PMA-differentiated THP-1 cells robustly induced hTLR8 (∼11-fold) and hTLR7 (∼3-fold) mRNA (Fig. 2,a). More importantly, IFN-γ priming of the cells restored the sequence-specific TNF-α response observed in PBMCs (Fig. 2,b and data not shown). B-406-AS, Lamin-AS, and STAT-2AS significantly induced TNF-α, but not B-406-S and si9.2-S, in agreement with the findings in PBMCs (compare Figs. 1,a and 2b, TNF-α). In fact, screening of all 14 sequences in the IFN-γ-primed THP-1 showed good agreement with the response profiles seen in PBMCs (data not shown), with a few exceptions. For example, GFP21-AS induced TNF-α in THP-1 but not in PBMCs, where it potently induced IFN-α (Fig. 1 a). This suggested that TNF-α production in the THP-1 model could also be indicative of sequences that preferentially induce high IFN-α in PBMCs, thus revealing a possible overlap of hTLR7 and hTLR8 signaling pathways in THP-1 cells.
hTLR7 is required for sequence-specific sensing by macrophages
To investigate the respective contributions of hTLR7 and hTLR8 in sequence-specific sensing in IFN-γ-primed THP-1 cells, RNA interference was used to selectively down-regulate hTLR7 or hTLR8 before priming. Fig. 3,a shows that significant knockdown of mRNA expression was achieved for both hTLR7 and 8. Attenuation of IFN-γ-driven hTLR8 up-regulation by siTLR8 functionally resulted in a 4-fold decrease of TNF-α production with the hTLR8 agonist 3M-002 (Fig. 3,b). siRNA silencing of hTLR7, but not 8, abrogated the response to loxoribine, a synthetic hTLR7 agonist. In contrast, silencing of both receptors had no effect on LPS signaling, a TLR4 ligand (Fig. 3 b). Interestingly, hTLR7 down-regulation also affected 3M-002 sensing (2.5-fold decrease, compare siTLR7 vs siGAPDH conditions). This is in agreement with some cross-reactivity of 3M-002 to hTLR7, as reported by the manufacturer at the concentration used (Invivogen).
The effect of down-regulating hTLR7 and hTLR8 was next assessed in the sensing of three ssRNA sequences (STAT-2AS, SA, and GFP21-AS). These were chosen on the basis of their different cytokine induction profiles in PBMCs (Fig. 1), where SA produced only a TNF-α response (thus indicating a possible preference for hTLR8) while GFP21-AS preferentially induced IFN-α (suggesting its preference for hTLR7), whereas STAT-2AS induced both cytokines. Unexpectedly, hTLR7 down-regulation by siTLR7 significantly affected sensing of all three ssRNAs (Fig. 3,c), demonstrating direct involvement of this receptor in the sensing of all types of immunostimulatory RNA oligonucleotides identified in this study. The sensing of SA, a putative hTLR8 ligand, was significantly decreased by hTLR7 down-regulation and this effect was more than that seen with down-regulation of hTLR8. Nevertheless, hTLR8 down-regulation by siTLR8 also affected sensing of ssRNA (Fig. 3 c, STAT-2AS), although to a lesser extent than with hTLR7 knockdown.
Variation of relative levels of endogenous hTLR7 to hTLR8 affects sequence-dependent sensing
Recently, possible interaction between hTLR7 and hTLR8 was inferred with the finding that both receptors could be immunoprecipitated following overexpression in HEK293 cells (27). A negative-regulatory role for hTLR8 on hTLR7 sensing of synthetic agonists was also suggested (27). To further investigate the possible impact of hTLR8 on the sequence-specific detection of ssRNA through hTLR7, we investigated how the modulation of expression levels of hTLR8 relative to hTLR7 related to ssRNA sensing, in the THP-1 model. Having shown that IFN-γ priming of PMA-differentiated THP-1 favored hTLR8 over hTLR7 expression (thus mimicking the balance observed in monocytes) (Fig. 2,a), we searched for a treatment resulting in favored hTLR7 expression compared with hTLR8 (thus modeling the balance seen in pDCs). In agreement with a previous report (26), IL-6 priming of PMA-differentiated THP-1 cells resulted in a preferential up-regulation of hTLR7 (∼6-fold) vs hTLR8 (∼2-fold) (Fig. 4,a), thereby mimicking the ratio of hTLR7 to hTLR8 observed in pDCs. To determine the effect of this on immunostimulation, we compared the activities of four ssRNAs sequences with distinct cytokine induction profiles in PBMCs, on IFN-γ- and IL-6-primed THP-1 cells. Although STAT-2AS strongly induced TNF-α in both systems in a dose-dependent manner, discrepancies were observed between si9.2-S, SA (that both have a similar amount of uridines, see Table I), and B-406-S (Fig. 4, b and c). In accordance with the preferential induction of hTLR7, the ability of si9.2-S to promote immunostimulation was significantly higher than that of SA and B-406-S in IL-6-primed cells (Fig. 4,c). This correlated with a dose-response induction of IFN-α with si9.2-S, whereas neither SA nor B-406-S potently induced IFN-α in PBMCs (Fig. 4,e). Conversely, SA was significantly more immunostimulatory than si9.2-S and B-406-S in IFN-γ-primed THP-1 cells and in PBMCs as measured by TNF-α production (Fig. 4, b and d).
Induction of hTLR7 and 8 further correlated to sequence-specific ssRNA sensing in primary CD14+ monocytes pretreated with IFN-γ (Fig. 5, a and b). Although no immunostimulation could be seen without IFN-γ pretreatment of CD14+ cells for SA and si9.2-S, pretreatment resulted in specific sensing of SA over si9.2-S (Fig. 5 b).
Interspecies conservation of ssRNA sensing in macrophages via TLR7
mTLR8 is not involved in the detection of ssRNAs in mouse macrophages, but rather this has been proposed to rely solely on mTLR7 from studies using TLR7−/− and TLR8−/− mice (11, 18). Our finding of a direct role for hTLR7 in sequence-specific sensing of ssRNAs in THP-1 macrophages, in conjunction with hTLR8, led us to investigate whether sequences that were immunostimulatory in humans were also immunostimulatory in mouse macrophages. In agreement with a conservation of sequence-specific sensing of ssRNAs between human and mouse macrophages (11, 18), all the sequences that induced IFN-α in human PBMCs activated NF-κB and induced TNF-α production in mouse macrophage RAW-ELAM cells, in a dose-dependent manner (Fig. 6 and data not shown). This sensing of ssRNAs was TLR7 dependent, as demonstrated by the lack of immunostimulation in TLR7−/− bone marrow macrophages (BMMs) (Fig. 6 b).
However, two ssRNA sequences inducing TNF-α but not IFN-α in human PBMCs (hence showing a possible preference for hTLR8) did not activate mTLR7 in mouse macrophages, even at high concentration (Fig. 6, see SA, and data not shown). These results are in accord with the findings by Heil et al. (Ref. 11 , see ssRNA42), and suggest that the discrepancies in ssRNA sensing between human and mouse macrophages (as shown with SA) rely on the lack of mTLR8 function, while TLR7 function and sequence-specific sensing is conserved between species.
Discussion
The potential of RNA oligonucleotides for use as therapeutic agents is currently under intense investigation. Several reports have implicated activation of hTLR7 and 8 by siRNA duplexes complexed with cationic liposomes in human blood cells (13, 14, 15, 16, 18). The induction of innate immunity was sequence dependent and was also observed with at least one of the two strands of the immunostimulatory duplex, when used separately (15, 16, 18).
Although investigating the mechanisms underlying the immune response invoked by short ssRNAs in human PBMCs, we observed differential cytokine induction profiles (Fig. 1,a). Although some sequences potently promoted both IFN-α and TNF-α induction (B-406-AS, Lamin-AS, STAT-2AS), others were restricted to induction of IFN-α (si9.2-S, GFP21-AS) at the concentration used. The addition of a UU motif within a predicted stem loop region of a nonstimulatory ssRNA (B-406-S giving SA) resulted in specific TNF-α induction independent of IFN-α, even at very high doses in PBMCs (Figs. 1,c and 4, d and e). Although this supports previous claims that RNA immunostimulation is directly dependent on the amount of uridine residues in human PBMCs (11, 16), we found that localization of the uridines within the ssRNA was also important (Fig. 1, b and c, compare B-406-S and SB). Assuming that IFN-α production is a readout of pDC activation and that of TNF-α monocyte activation, it is possible that some ssRNAs preferentially recruit hTLR7 (in pDCs) or hTLR8 (in monocytes) in a similar manner to that of synthetic immune response modifiers (9, 10, 16).
To further characterize the involvement of hTLR7 and hTLR8 in sequence-specific ssRNA sensing, we established a macrophage-like cell model (THP-1 cells) (25) reproducing functional detection of immunostimulatory RNA oligonucleotides (Fig. 2). TNF-α production by ssRNA-stimulated cells was only observed after IFN-γ priming of the cells, which induced both hTLR7 and hTLR8 expression—with a preference for hTLR8 (Fig. 2,a). As shown in Fig. 5, a and c, this was replicated in another monocytic/macrophage-like cell line (U937 cells, see STAT-2AS condition) (25), although the THP-1 model appeared to be more sensitive to ssRNAs (and yielded much higher TNF-α levels). The sequence-specific cytokine induction profile in THP-1 cell model paralleled TNF-α production observed in PBMCs (Figs. 1,a and 2,b). One sequence inducing IFN-α but not TNF-α in PBMCs (GFP21-AS) was found to induce TNF-α in this model. This is likely related to the higher concentration of ssRNAs used in THP-1 compared with PBMCs (Figs. 1,a and 2,b), as observed with si9.2-S which induces TNF-α in PBMCs when used at doses higher than 200 nM (Fig. 4 d).
To directly study the involvement of hTLR7 and 8 in the response observed in these macrophage-like cells, we used RNA interference to knockdown hTLR7 or hTLR8 in THP-1 cells. This altered IFN-γ-induced hTLR7 and hTLR8 up-regulation (Fig. 3,a) and resulted in a significant decrease of TNF-α production by synthetic hTLR7/8 ligands, but not the TLR4 ligand LPS (Fig. 3,b). Unexpectedly, down-regulation of both hTLR8 and hTLR7 functionally altered ssRNA sensing (Fig. 3,c, STAT-2AS). The down-regulation of hTLR7 also resulted in a stronger impairment of sensing of all three ssRNAs tested when compared with hTLR8 RNA interference (Fig. 3,c), probably as a result of more efficient siTLR7 inhibition of hTLR7 induction by IFN-γ (Fig. 3 a).
Taken together, these results demonstrate a direct involvement of hTLR7 in the induction of TNF-α by ssRNAs in the macrophage-like THP-1 cell line. Consistent with this, sequence-dependent ssRNA detection by murine BMMs was also dependent on TLR7, as illustrated by the ablation of the TNF-α response in TLR7−/− BMMs (Fig. 6,b and Refs. 11 and 18). Furthermore, all inducers of IFN-α in human PBMCs were also immunostimulatory in mouse macrophages. It should be noted however, that the sensing of a sequence favored by an overexpression of hTLR8 vs hTLR7 in THP-1 cells and CD14+ monocytes (Figs. 4, b and d, and 5,b, SA) and which did not induce IFN-α in PBMCs, is not conserved in mouse BMMs (Fig. 6, a and b, SA).
Based on our findings, we propose a model where TLR7 is involved in sequence-specific sensing of RNA oligonucleotides in both mouse and human macrophages; nevertheless, sensing of ssRNAs by hTLR7 would be complemented by hTLR8 in human but not in mouse macrophages. In cytokine-primed THP-1 cells, the relative expression of hTLR7 to hTLR8 modulates the sequence-dependent sensing of ssRNAs with an identical amount of uridine residues (Fig. 4, b and c, and Table I, compare SA and si9.2-S). We also show that primary CD14+ monocytes produce TNF-α in response to some ssRNAs, and that sequence-specific sensing is enhanced with IFN-γ priming (Fig. 5,b, compare STAT-2AS, SA and si9.2-S); of note, we did not detect IFN-α production in these cells (data not shown). Enhanced TNF-α levels in CD14+ cells correlated with an induction of both hTLR7 and 8 mRNA levels, with a predominance of hTLR8 (Fig. 5,a). The observation that all monocytic/macrophage cells used in this study induced both hTLR7 and 8 in response to IFN-γ priming, which correlated with sequence-specific sensing of ssRNAs, indicates that expression of the two TLRs is functionally related in human macrophages. This adds credence to a role for hTLR7, together with hTLR8 in ssRNA sensing in human macrophages. Indeed, IFN-γ inducibility of hTLR7 and 8 is most likely relevant to a complex feedback of cytokines produced by stimulated PBMCs. NK cell activation by hTLR7/8 agonists has recently been shown to be dependent on the production of IL-12 by activated macrophages and to result in IFN-γ production (29, 30). It is therefore likely that the hTLR7/8 responsiveness to IFN-γ in monocytes/macrophages observed here is part of a loopback mechanism, further increasing sensing of hTLR7/8 agonists by macrophages. In support of this is the observation that the levels of TNF-α from unprimed CD14+ cells treated with ssRNAs were >10-fold lower than that observed with PBMCs, even though the number of monocytes was ∼2- to 4-fold higher (compare Figs. 5,b and 1 a, STAT-2AS).
The finding of a requirement for hTLR7 in ssRNA sensing by THP-1 cells was unexpected given the prevalence of hTLR8 induction by IFN-γ in this cell type. Furthermore, the previous demonstration that transient overexpression of hTLR8 in HEK293 and U20S cells reconstitutes responsiveness to ssRNAs via activation of NF-κB (11, 31) indicates that hTLR8 does not require hTLR7 for signaling when expressed artificially. The discrepancies between these results and the data presented here are likely related to overexpression of TLRs in these studies and the saturation doses of ssRNAs used (∼3-fold higher than the highest dose used here). In addition, the data from these studies demonstrated that hTLR8 senses uridine-containing ssRNAs (compare ssRNA 40, 41, and 42; Ref. 11), but did not provide evidence that variation in the localization of the uridine residues was related to varied immunostimulation. Hence, these results did not specifically show a sequence-specific response of hTLR8 to ssRNA. Although we implicate hTLR8 in ssRNA sensing in THP-1 cells (Fig. 3 c, STAT-2AS), we suggest it complements hTLR7 activity and thereby broadens the repertoire of ssRNA sequence sensing. This is supported by the identification of a ssRNA (SA) which signals preferentially through hTLR8 but also engages hTLR7 and is not immunostimulatory in mouse where mTLR8 is not functionally involved in ssRNA sensing (11).
In conclusion, we suggest that the function of hTLR8 in ssRNA sensing broadens the range of immunostimulatory sequences and related cytokine induction profile induced by hTLR7 engagement in human macrophages. Whether the regulation of hTLR7 ssRNA sensing by hTLR8 is related to a direct binding of the two receptors with the ssRNAs in a similar fashion to that of TLR9 dimerization (32) remains to be determined. In this work, we also establish a novel cell model of ssRNA sensing by hTLR7/8 in human macrophages, relying on IFN-γ-activated THP-1 cells. Our data indicate that in conjunction with the murine RAW cell model previously reported (13), prediction and screening of immunostimulatory ssRNAs that have potential in therapy, such as siRNAs, can be accurately done using these two established cell lines.
Acknowledgment
We are grateful to N. Skinner for her assistance in the isolation of CD14+ monocytes.
Disclosures
Dr. Mark Behlke is employed by Integrated DNA Technologies Inc. (IDT), which offers oligonucleotides for sale similar to some of the compounds described in the manuscript. IDT is however not a publicly traded company, and he does not own any shares/equity in IDT.
Footnotes
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work was supported by National Institutes of Health Grants RO1 A134039 and PO1 CA62220 (to B.R.G.W.).
Abbreviations used in this paper: hTLR7/8, human TLR7/8; pDC, plasmacytoid dendritic cell; mTLR7/8, mouse TLR7/8; siRNA, small interfering RNA; BMM, bone marrow macrophage; DOTAP, N-[1-(2,3-Dioleoyloxy)propyl]-N,N,Ntrimethylammonium methylsulfate.