TLR8 Senses Staphylococcus aureus RNA in Human Primary Monocytes and Macrophages and Induces IFN-β Production via a TAK1–IKKβ–IRF5 Signaling Pathway

Staphylococcus aureus can act as a peaceful colonizer of the skin and the nostrils or as an aggressive pathogen causing invasive diseases and sepsis. Intracellular survival of S. aureus, abscess formation, as well as the emergence of methicillin-resistant strains complicate the treatment of serious infections (1). S. aureus also produces virulence factors such as hemolytic and leucolytic toxins and C-targeting factors that contribute to immune evasion (2). TLR2 is a primary pattern recognition receptor (PRR) for sensing of S. aureus by immune cells and mediates resistance of mice against experimental S. aureus infection (3). S. aureus deficient in lipoprotein synthesis (Δlgt) does not activate TLR2 and elicits reduced proinflammatory responses in human cell lines (4). Children and teenagers with MyD88- or IL-1R–associated kinase 4 deficiency are at risk for infections with pyogenic bacteria, in particular S. aureus, Streptococcus pneumonia, and Pseudomonas aeruginosa, whereas their resistance against infection with other pathogens is normal (5, 6). This indicates a particular importance of TLR2 and/or IL-1Rs for resistance against S. aureus infection in young humans.

Type I IFNs are classical antiviral cytokines, but they are also induced by intracellular and extracellular bacteria. The impact of the main type I IFNs (IFN-α and IFN-β) on bacterial infections is less clear and spans from enhanced innate and cell-mediated immunity to immune suppression and dysregulation, which may contribute to the progression of septic shock (7). The predominant pathway of IFN-β induction by Gram-negative bacteria is by LPS-mediated activation of endosomal TLR4 signaling via the Toll/IL-1R domain–containing adapter inducing IFN-β–IFN regulatory factor (IRF)3 pathway (8, 9). For Gram-positive bacteria there may not be a single predominant mechanism for IFN-β induction, as multiple pathogen-associated molecular patterns (PAMPs) and cell host receptors have been implicated in different model systems. In murine phagocytes, recognition of bacterial RNA and DNA appears as central for IFN-β production (10–12), and mouse TLR13 was recently identified as a sensor of RNA of microbial origin (13, 14). Induction of type I IFNs by S. aureus has been examined in different human and murine cell types with different conclusions regarding the molecular mechanisms involved. S. aureus PAMPs suggested to be responsible for IFN-β induction include staphylococcal protein A, DNA, RNA, and lipoteichoic acid, and TLR2, TLR7, TLR9, and cytosolic PRRs were implicated (15–22).

The aim of the present study was to examine the role of TLR2 and other PRRs for S. aureus–induced IFN-β production in human primary monocytes and monocyte-derived macrophages (MDMs). Unexpectedly, we found that TLR2 activation could suppress the S. aureus–induced production of IFN-β. In contrast, induction of IFN-β was triggered by S. aureus RNA, which activated a TLR8–IRF5 signaling axis in a TGF-β–activated kinase 1 (TAK1)– and IκB kinase (IKK)β–dependent fashion. In this study, we establish TLR8 as a second MyD88-dependent PRR of S. aureus in human primary monocytes and MDMs and show that it is essential for the induction of IFN-β production by whole bacteria via a recently identified IKKβ–IRF5 activation pathway. We also demonstrate a cross-regulatory function of TLR2 in TLR8–IRF5 signaling.

Concentrations of IFN-β were determined with the VeriKine-HS human IFN-β serum ELISA kit (PBL Assay Science, Piscataway, NJ) typically with no dilution or 1:2 dilutions of the culture supernatants. BioPlex assays were from Bio-Rad and were analyzed as per the manufacturer’s instructions, with dilutions of the supernatants ranging from none to 200, depending on the cytokine to be examined. Escherichia coli bioparticles were of the rough K12 strain (Invitrogen). The following TLR ligands were from InvivoGen: LPS from the E. coli K12 strain, FSL-1, Pam3Cys, R837, CL75, polyinosinic-polycytidylic acid [poly(I:C)] and polyuridylic acid (pU). Lipofectamine 2000 (L2K) was from Invitrogen, and poly-l-arginine (pL-Arg) was from Sigma-Aldrich. RNAse A was from Qiagen. The IKKβ inhibitor BI605906 was provided by Prof. Sir Philip Cohen (University of Dundee, Dundee, U.K.), and IKKII–VIII and the TAK1 inhibitor 5Z-7-oxozeaenol was from Calbiochem/Merck Millipore (Darmstadt, Germany). The TAK1 inhibitor NG-25 was from MedChem Express (Monmouth Junction, NJ).

S. aureus 113 wild-type (wt) strain, its isogenic 113 Δlgt mutant, and the pRBlgt-reconstituted 113 Δlgt strain were provided by Prof. Friedrich Göetz (University of Tübingen, Tübingen, Germany). The Newman and Cowan strains were provided by Prof. Timothy Foster (Trinity College, Dublin, Ireland), and the Wood-46 strain was from the American Type Culture Collection (10832). The bacteria were grown on tryptic soy agar plates that were supplemented with 10 μg/ml erythromycin or kanamycin for the Δlgt mutant or the pRBlgt reconstituted strains, respectively. For preparation of bacteria, colonies were picked and grown in 5 ml tryptic soy broth during vigorously shaking at 37°C overnight (12–18 h) for use in infection experiments. To prepare heat-killed (HK) bacteria, a preculture was diluted 1:100 to 1:200 in 50–100 ml tryptic soy broth in 500–1000 ml culture flasks and grown in a shaking incubator to the exponential phase (∼4 h), stationary phase (∼12 h), or decline phase (20–24 h), as appropriate. For heat killing of the bacteria, bacteria were spun down, resuspended in PBS, incubated at 80°C for 30 min, and finally washed once with PBS. For quantification of bacteria by OD measurements, a standard curve was generated with serial dilutions of HK bacteria that had been quantified by manual counting in a Bürker chamber. Fluorescent labeling of the HK bacteria was done using Alexa Fluor 488–succinimidyl ester (Invitrogen). This reagent was dissolved in DMSO and immediately added to a solution of 2 × 10¹⁰ S. aureus/ml in NaHCO₃ (167 μM, pH 8.3), yielding a final concentration of 1 mg/ml dye. Incubation with agitation was done for 1 h at room temperature, and the labeled bacteria were washed twice with 500 μl PBS and counted. Preparation of crude bacterial lysate was done by a previous described protocol (4) with some modifications. Glass particles (0.1 mm, Sigma-Aldrich) were preheated at 200°C for 4 h to eliminate potential TLR ligand contaminants. The particles were added to 2 × 10¹⁰ S. aureus 113 Δlgt (HK, exponential growth phase) in ice-cold PBS and run in four cycles on a Precellys 24 bead-beater (Bertin Technologies, Montigny-le-Bretonneux, France) with chilling on ice between each cycle. The glass particles and intact bacteria were spun down and the crude lysate supernatants were added to new tubes for storage at −80°C.

Fresh blood, serum, and buffy coats were acquired from healthy volunteers under informed written consent approved by the Regional Committee for Medical and Health Research Ethics (REC Central, Norway, no. 2009/2245). Human PBMCs were isolated from buffy coats using Lymphoprep (Axis-Shield) as described by the manufacturer, and monocytes were purified by adherence in culture plates and maintained in RPMI 1640 (Life Technologies) supplemented with 10% pooled human serum (HS). TLR2 blocking was done by 30 min pretreatment with the anti-mouse/human TLR2 mAb clone T2.5 (no. HM1054, Hycult Biotech) at 5 μg/ml, and mouse IgG1 mAb (R&D Systems) served as isotype control. Poly(I:C), pU, and crude S. aureus lysate, with or without RNAse A (Qiagen) treatment (2 μl/ml, 1 h at 37°C), was precomplexed with pL-Arg at a 1:1 ratio (w/w) in Opti-MEM (Invitrogen), or with 2.5 μl L2K/μg RNA in Opti-MEM for transfection. For infection of monocytes and macrophages, live bacteria from overnight cultures with or without the bacterial culture media were diluted in RPMI 1640 and incubated for 1 h at room temperature before addition to the cells. Extracellular bacteria were killed after 1 h by addition of gentamicin to 100 μg/ml. For quantitative real-time PCR (Q-PCR) analysis, the cells were lysed after a total infection time of 3–4 h, and cell culture supernatants were harvested after 5–6 h of infection.

Macrophages were derived from monocytes by differentiation in RPMI 1640 with 30% pooled HS for 5–6 d. Medium was replaced with RPMI 1640 containing 10% serum for infection, stimulation, or siRNA treatment. A pool of four individual ON-TARGETplus siRNAs (Dharmacon) was transfected using siLentFect (Bio-Rad), yielding a final concentration of 5 nM siRNA. The transfection was repeated after 3 d, and the silenced MDMs were infected with live S. aureus for 3 h. RNA was isolated with an RNeasy kit including DNAse treatment (Qiagen), cDNA was transcribed with a Maxima cDNA synthesis kit (Thermo Fisher Scientific), and relative quantification by Q-PCR was done with StepOnePlus using TaqMan probes (Life Technologies) and Perfecta qPCR FastMix from Quanta. Probes used were: IFN-β, Hs01077958_s1; TNF, Hs00174128_m1; TLR7, Hs00152971_m1; TLR8, Hs00607866_mH; IRF5, Hs00158114_m1; stimulator of IFN genes (STING), Hs00736958_m1; and TBP, Hs00427620_m1. TBP served as endogenous control, and relative expression in monocytes was calculated as fold induction by stimulation or infection. For siRNA-treated macrophages, the level of gene expression by infected cells were normalized against noninfected cells treated with each of the respective siRNAs to correct for potential background differences.

The whole blood experiments were performed basically as described (23). Venous blood was drawn into polypropylene tubes with lepirudin (Refludan) anticoagulant (50 μg/ml final concentration). The blood was quickly transferred to 1.8-ml round-bottom cryotubes (Nunc) containing HK S. aureus and FSL-1 and then incubated on a tube roller at 37°C. Samples for cytokine analyses were centrifuged for collection of plasma, and samples for flow cytometry analyses were fixed immediately with 0.5% paraformaldehyde (PFA) for 5 min. For analysis of CD11b expression, blood cells stimulated with HK bacteria and/or FLS-1 for 15 min were stained with anti-CD11b PE and anti-CD14 FITC (BD Biosciences) and EasyLyse erythrocyte-lysing reagent (Dako). Flow cytometry was performed on a BD LSR II using side scatter and CD14 to gate for neutrophils and monocytes, and CD11b expression was determined by the median fluorescence intensity (PE) of the total cell distribution of monocytes and granulocytes. Phagocytosis of Alexa Fluor 488–labeled HK S. aureus was analyzed with the Phagotest kit (BD Biosystems) according to the manufacturer’s instructions. In brief, full blood was incubated with bacteria with or without FSL-1 at 37°C for 30 min, and control tube with S. aureus was kept on ice. Phagocytosis was stopped by placing tubes on ice and immediately adding ice-cold quenching solution. This was followed by washing with cold washing buffer, lysis of erythrocytes, and fixation of leukocytes with lysis solution, and the DNA stain was finally added to exclude artifacts of aggregated bacteria and cells. Monocytes and granulocytes were discriminated by forward scatter/side scatter, and a phagocytic index was calculated as: mean fluorescence intensity (Alexa Fluor 488) of cells containing bacteria × fraction of cells containing bacteria.

Cells were adhered in six-well plates, treated and lysed in 150 μl lysis buffer (20 mM Tris-HCl, 1 mM EDTA, 1 mM EGTA, 137 mM NaCl, 1% Triton X-100, 1 mM sodium deoxycholate, 10% glycerol, 1 mM Na₃VO₄, 50 mM NaF, and Complete protease inhibitor [Roche]). PAGE was performed with the NuPAGE system (Life Technologies) per the manufacturer’s recommendations. Immunoblotting was performed with the iBlot system (Life Technologies) per the manufacturer’s recommendations. After blotting, nitrocellulose membranes were briefly rinsed with dH₂0 and blocked with 5% BSA dissolved in TBST for 1 h. All incubations with primary Abs were done at 4°C overnight, and all incubations with secondary HRP-linked Abs (Dako, nos. P0399 and P0447) were done at room temperature for 1 h. Blots were developed with SuperSignal West Femto substrate (Pierce) and imaged with a Li-Cor Odyssey Fc system. Abs used for Western blots in this study were: p38 (Cell Signaling Technology, no. 9212), phospho-p38 (Cell Signaling Technology, no. 9211), p44/42 (ERK; Cell Signaling Technology, no. 4695), phospho-p44/42 (ERK, Cell Signaling Technology, no. 4370), JNK (Cell Signaling Technology, no. 9252), phospho-JNK (Cell Signaling Technology, no. 4668), TANK-binding kinase 1 (TBK1; Cell Signaling Technology, no. 3504), phospho-TBK1 (Cell Signaling Technology, no. 5483), phospho-IKKα/β (Cell Signaling Technology, no. 2697), TAB1 (Cell Signaling Technology, no. 3226), phospho-TAB1 (Millipore, no. 06-1334), phospho-p105 (Cell Signaling Technology, no. 4806), IkBα (Cell Signaling Technology, no. 4812), GAPDH (Abcam, no. ab8245), TLR8 (Cell Signaling Technology, no. 11886), and IRF5 (Cell Signaling Technology, no. 13496).

Immunofluorescence labeling was done as described (9) with minor changes. PBMCs were seeded in 96-well glass bottom plates (no. P96-1.5H-N, In Vitro Scientific, Sunnyvale, CA) that were precoated with HS for 1–2 h. Nonadherent cells were removed by washing with 3× HBSS. For intracellular staining the cells were fixed with ice-cold 2% PFA in PBS for 15 min on ice and washed three times with room-tempered PBS. Permeabilization was done with PEM buffer (100 mM K-Pipes [pH 6.8], 5 mM EGTA, 2 mM MgCl₂, 0.05% saponin) for 10 min, quenched of free aldehyde groups in 50 mM NH₄Cl in PBS with 0.05% saponin (PBS-S) for 5 min, and blocked with 20% HS in PBS-S for 20 min. After a single wash with 1% HS in PBS-S, the cells were incubated with primary Ab (2 μg/ml) in with 1% HS in PBS-S overnight at 4°C. Cells were washed twice with room-tempered PBS-S and once with 1% HS in PBS-S, and incubated with highly cross-adsorbed Alexa Fluor 488– or 647–labeled secondary Abs (Invitrogen) at 2 μg/ml for 30 min. Subsequently, the cells were washed with 3× PBS-S, postfixed with 4% PFA at room temperature, and washed once with PBS-S. Nuclei were stained with Hoechst 3342 (200 ng/ml) in PBS-S. The following Abs were used (typically 2–10 μg/ml or 1:100- to 1:200-fold dilution): anti-human TLR8 (Novus Biologicals, no. NBP1-77203), rabbit IgG control (Novus Biologicals, no. NB810-56910), anti-human IRF5 mAb (Abcam, no. 10T1), anti-human IRF5 mAb (Abcam, no. ab124792), anti-human IRF5 (Santa Cruz Biotechnology, no. D10), anti-human IRF5 (Sigma-Aldrich, no. HPA046700), anti-human IRF3 XP mAb (Cell Signaling Technology, no. 11904), anti-human IRF3 mAb (D83B9 mAb) (Cell Signaling Technology, no. 4302), anti-human IRF3 (Santa Cruz Biotechnology, FL425, no. sc9082), anti-human p65 XP mAb (Cell Signaling Technology, no. 8242), anti-human p65 A (Santa Cruz Biotechnology, no. sc-109), anti-human IRF1 XP mAb (Cell Signaling Technology, no. 8478), anti-human IRF7 (Cell Signaling Technology, no. 4920; EPR4718, Abcam, no. ab109255; H-246, Santa Cruz Biotechnology, no. sc-9083), anti–phospho-IRF7 (Cell Signaling Technology, no. 5184), anti-human IRF8 (Santa Cruz Biotechnology, C-19, no. sc-6058; Sigma-Aldrich, no. HPA00253; Cell Signaling Technology, no. 5628), normal goat IgG (Santa Cruz Biotechnology), and rabbit IgG XP control mAb (Cell Signaling Technology). Automated imaging was done with the Scan^R system (Olympus) using a ×20 objective, up to 1 s exposure time, and ∼100 frames were captured for each well (∼5,000–15,000 cells) performed in duplicates or triplicates. Automated image analysis was done with the Scan^R software v1.3.0.3. Confocal images were captured with a Zeiss LSM 510 META scanning unit, and a 1.4 numerical aperture ×63 objective.

Statistical analyses were done on data merged from independent experiments, as indicated in the figures or the figure legends. For analysis, the data were log₂ transformed to generate Gaussian distributions, and analyses were performed with GraphPad Prism v5.03. Figures without statistics show a single representative experiment out of at least three independent experiments.

To clarify the role of TLR2 ligands in S. aureus–induced IFN-β production, we infected monocytes with live S. aureus 113wt strain and its isogenic mutant 113 Δlgt. The mutant strain is deficient in mature lipoprotein production and fails to activate TLR2, resulting in impaired immune activation by entire bacteria (4, 24). To examine the total immunostimulatory capacity of the different strains, we included both the bacteria and the supernatants from the bacterial cultures, and the plasmid-reconstituted 113 Δlgt strain (pRBlgt) served as a genotype/phenotype control (Fig. 1A). Unexpectedly, the Δlgt strain induced higher levels of IFN-β than did the wt and pRBlgt strains, suggesting that S. aureus lipoproteins inhibit IFN-β induction by the bacteria. To examine the possible contribution of factors in the bacterial culture supernatant for this phenomenon, the wt and Δlgt supernatants were swapped (Fig. 1B). This resulted in increased IFN-β induction by the wt strain and concomitant increased IFN-β induction by the mutant strain, thus reaching intermediate IFN-β levels compared with the condition in Fig. 1A. This indicates that the S. aureus 113 wt strain inhibits IFN-β induction by lipoproteins that are released into the bacterial culture media during growth, as well as by lipoproteins in the bacterial cell wall. Release of lipoproteins by the 113 wt strain during growth has been demonstrated previously (4), and with a TLR2–HEK293–NF-κB reporter assay we identified TLR2 ligands in the culture supernatant of the 113 wt strain, but not the Δlgt strain (not shown). To examine whether TLR2 ligands of S. aureus antagonize the induction of IFN-β during infection, we used a TLR2-blocking Ab (Fig. 1C). In the presence of bacterial culture media, the TLR2-blocing Ab increased the production of IFN-β by monocytes upon infection of all the S. aureus strains examined, except for the 113 Δlgt strain. In the absence of bacterial culture media, TLR2 blocking significantly increased the IFN-β production by the pathogenic S. aureus isolates Newman and Cowan, with a similar tendency for the 113 wt and Wood46 strains. Collectively, this suggests that both lipoproteins released into the bacterial culture media and in the bacterial cell wall can antagonize the IFN-β production from monocytes via activation of TLR2. We further examined the role of TLR2 in the induction of other cytokines using multiplex ELISA. Secretion of IL-6, IL-8, IL-1α, and IL-12-p40 by monocytes in response to S. aureus infection was partially lipoprotein-dependent, whereas secretion IL-1β and IL-18 was independent of lipoproteins (Supplemental Fig. 1). Lipoproteins suppressed IL-12p70 secretion to a similar degree as IFN-β, suggesting that these two cytokines share a regulatory mechanism (Supplemental Fig. 1).

To simplify our model system, we examined whether the defined synthetic TLR2/6 ligand FSL-1 alone would antagonize IFN-β induction by HK S. aureus Δlgt, E. coli, and E. coli LPS (Fig. 2A). TLR2 costimulation with FSL-1 suppressed IFN-β induction by HK S. aureus, but not HK E. coli and E. coli LPS. This confirmed that TLR2 activation blocks S. aureus–induced IFN-β production, whereas it does not affect TLR4–TRIF signaling. S. aureus lipopeptides can be both diacylated and triacylated (25), and synthetic triacylated TLR1/2 ligand Pam3Cys also inhibited the IFN-β induction by HK S. aureus in monocytes (not shown).

We further examined the impact of the timing of TLR2 ligand administration relative to S. aureus and found that maximum inhibition of IFN-β induction was achieved when FSL-1 was added before or at the same time as the bacteria, although the inhibitory effect was gradually reduced when FSL-1 was added 15, 30, and 60 min later (Fig. 2B). Thus, the TLR2 inhibitory effect is limited to an early time frame of S. aureus sensing by monocytes.

To examine whether the inhibitory effect of TLR2 also is important in the more complex physiological environment of blood monocytes, we employed a lepirudin anticoagulated human whole blood model (ex vivo) that enables crosstalk between all blood cells and most of the plasma cascades, including a C system that is functionally active under physiological conditions (23). TLR2 costimulation fully suppressed S. aureus–induced IFN-β production in whole blood (Fig. 2C), confirming the validity of our PBMC-based model. Cytokine induction by HK S. aureus is to a large extent dependent on phagocytosis and bacterial degradation (20, 26). The TLR2 effect could thus possibly be explained by inhibition of phagocytosis. However, we found that TLR2 costimulation did not inhibit phagocytosis, but instead significantly increased the uptake of HK S. aureus by monocytes in whole blood (Fig. 2E). The mechanism for the enhanced phagocytosis is likely TLR2-mediated activation of the CR3 (CD11b/CD18), as S. aureus phagocytosis in the whole blood model is strongly C-dependent (27) and the CD11b level on both monocytes and granulocytes increased strongly upon TLR2 stimulation (Fig. 2F).

Because bacterial nucleic acids are implicated in type I IFN induction in various model systems, we examined the importance of RNA in crude lysate of the HK S. aureus Δlgt strain. Bacteria were mechanically disrupted, treated with RNAse A, which cleaves ssRNA, and the lysate was mixed with pL-Arg for delivery of nucleic acid to the monocyte endosomal compartment (28) (Fig. 3). RNAse treatment eliminated the induction of IFN-β, IL-12p40, and IL-12p70, and it strongly reduced the release of IL-1α, IL-1β, and IL-18. In contrast, the levels of IL-6 and IL-8 were not affected by RNAse A (Fig. 3). This demonstrates that in lysate of HK S. aureus deprived of TLR2 ligands, ssRNA was a dominant PAMP for the induction of IFN-β, IL-1, IL-12, and IL-18. Still, other PAMPs distinct from RNA and lipoproteins were apparently dominating for IL-6 and IL-8 induction by these crude lysates.

To clarify the mechanism of how S. aureus RNA induces IFN-β in monocytes, we compared the response by HK S. aureus Δlgt with the synthetic TLR8 ligand pU and the dsRNA TLR3 ligand poly(I:C). The RNA was delivered to the endosomal or the cytosolic compartment of monocytes by complexation with pL-Arg or transfection with L2K, respectively, as previously shown (28). Stimulation with S. aureus, pU/pL-Arg, and poly(I:C)/L2K induced IFN-β production, whereas stimulation with poly(I:C)/pL-Arg and pU/L2K did not (Fig. 4A). This implies that pU and poly(I:C) induce IFN-β from the endosomal and the cytosolic compartments, respectively, and is consistent with high TLR8 and low TLR3 levels in monocytes (29). Costimulation with FSL-1 strongly suppressed IFN-β induction by HK S. aureus and pU, but not poly(I:C). TLR2 activation thus interferes with TLR8 signaling, but not cytosolic dsRNA sensing or TLR4–TRIF signaling (Fig. 2A). Moreover, induction of IL-12p70 was solely induced by S. aureus and pU in endosomes and was also inhibited by TLR2 activation (Fig. 4B). GU- and U-rich ssRNA stimulate murine TLR7 and human TLR8 (30), and the induction of IL-12p70 is a well-described characteristic of human TLR8 (28, 31). Thus, the regulation of IFN-β and IL-12p70 production by monocytes is similar for S. aureus and a defined TLR8 ligand, including its inhibition by TLR2 activation. Additional correlative evidence for the involvement of TLR8 in the sensing of S. aureus was found using the TLR8-specific imidazoquinoline agonist CL75, as CL75-mediated IFN-β induction was antagonized by TLR2 costimulation (Fig. 4C). In contrast, IFN-β induction by the TLR7 specific ligand R837 was not antagonized by TLR2 costimulation (Fig. 4D), arguing against a role of TLR7 in the IFN-β induction by S. aureus. TLR8 immunofluorescence indicated that TLR8 was recruited to S. aureus phagosomes (Fig. 4E). Quantification using a high-content screening system suggested that ∼9% of the S. aureus phagosomes stained positive for TLR8 1 h after exposure to bacteria (Fig. 4F).

The correlative data suggested that TLR8 was responsible for S. aureus–induced IFN-β production. Western blot analysis demonstrated that although S. aureus and TLR2 ligands activated TBK1 phosphorylation, IRF3 was not activated (not shown). Furthermore, preliminary data suggested a possible involvement of IRF5. To clarify the mechanism, we performed gene silencing of MDMs by transient transfection with siRNA targeting TLR7, TLR8, IRF5, and STING. All targets were efficiently and specifically silenced on the mRNA level after sequential transfections (Supplemental Fig. 2A). Clear knockdown of TLR8 and IRF5 was also seen at the protein level (Supplemental Fig. 2B). However, two TLR8 bands of ∼100 kDa were not completely eliminated even several days after mRNA knockdown. These bands probably correspond to N-terminal fragments of TLR8 after cleavage in endosomes and can be a functional PRR as recently shown (32). MDMs showed similar responses as monocytes, as costimulation with synthetic TLR2 ligand or S. aureus lipoproteins strongly antagonized IFN-β induction by S. aureus and synthetic TLR8 ligand (not shown). Infection of silenced MDMs with live S. aureus Δlgt demonstrated that induction of IFN-β was dependent on TLR8 and IRF5, but not on TLR7 or STING (Fig. 5). Induction of TNF followed a similar trend as IFN-β upon TLR8 and IRF5 silencing, but it was not as strongly affected and reached statistically significance only for the TLR7 plus TLR8 combined knockdown. Collectively, this indicates an essential role of a TLR8–IRF5 pathway in the induction of IFN-β by S. aureus, which also may contribute to S. aureus–induced TNF production by monocytes.

To further examine IRF5 activation, we established a quantitative immunofluorescence method of transcription factor nuclear accumulation using high-content screening (Scan^R) (Supplemental Fig. 3). Monocytes were stimulated with HK S. aureus and TLR8 ligands, and nuclear accumulation of total p65 (NF-κB/RelA), IRF3 (Supplemental Fig. 3), and IRF5 (Fig. 6) was examined. The level of nuclear IRF5 was low in resting cells (Fig. 6A) and was strongly increased following pU/pL-Arg stimulation (Fig. 6B) and phagocytosis of HK S. aureus Δlgt (Fig. 6D). FSL-1 costimulation clearly reduced the nuclear accumulation of IRF5 induced by both stimuli (Fig. 6C, 6E), which thus correlates with suppressed IFN-β and IL-12 induction. Moreover, when S. aureus was heat inactivated during the stationary growth phase, the bacteria were markedly less potent as IFN-β inducers (not shown). The stationary phase S. aureus did not induce IRF5 nuclear accumulation (Fig. 5F), thus again demonstrating a correlation of S. aureus–induced IFN-β production and IRF5 nuclear accumulation in monocytes. Around 25% of the monocytes that had phagocytosed HK S. aureus Δlgt stained positive for nuclear IRF5 (Fig. 5G). The frequency was strongly reduced by costimulation with FSL-1, as well as when HK S. aureus from the stationary growth phase was used. Moreover, cells that did not phagocytosed bacteria, that is, “S. aureus bystanders,” did not have increased IRF5 nuclear accumulation. Thus, IRF5 nuclear accumulation was dependent on S. aureus phagocytosis and not a result of paracrine signals. Correlation of IRF5 translocation and phagocytosis was also seen in a whole-well overview with S. aureus phagocytosis and IRF5 nuclear staining being strongest around the well center (not shown). pU/pL-Arg stimulation activated IRF5 nuclear accumulation to a similar degree as S. aureus uptake, and was also suppressed by FSL-1 costimulation, whereas FSL-1 or LPS alone did not activate IRF5 (Fig. 5H). In contrast, nuclear accumulation of p65 was seen with ligands for all three TLRs examined (Fig. 5I). Only LPS activated IRF3 nuclear translocation (Fig. 5J), which is consistent with the IRF3 phosphorylation pattern (not shown). IRF1 was constitutively localized to the nuclei in monocytes, whereas IRF7 and IRF8 Abs gave no specific staining (not shown). We conclude that S. aureus induces IRF5 nuclear accumulation in monocytes as a consequence of TLR8 activation and is blocked by TLR2 signaling.

We further examined the requirement of central signaling components in the MyD88 pathway for TLR8-mediated IRF5 and p65 activation and TLR2-induced p65 activation. We quantified IRF5/p65 nuclear staining by two-color immunofluorescence (Fig. 7A–C). We then used well-characterized chemical inhibitors of central signaling kinases to block monocyte signaling. To minimize possible problems of toxicity and secondary effects of the inhibitors, we chose CL75 as TLR8 ligand, as CL75 elicits more rapid cytokine induction than do pU/pL-Arg and S. aureus and thus limits the required incubation time with the inhibitors. Two structurally nonrelated inhibitors of TAK1 (5Z-7-oxozeaenol and NG-25) and IKKβ (IKKII-VIII and BI605906) were given as a 30-min pretreatment, and their effects on TLR8- and TLR2-mediated nuclear accumulation of IRF5 and/or p65 were examined (Fig. 7D–F). TLR8-induced IRF5 nuclear accumulation was dependent on both TAK1 and IKKβ (Fig. 7D). In contrast, p65 nuclear accumulation following TLR8 and TLR2 stimulation was only dependent on IKKβ and not on TAK1 (Fig. 7E, 7F).

Western blot analysis of MAPKs and other signaling intermediates was performed to verify the specificity of the inhibitors and to dissect further details of TLR2 and TLR8 signaling (Supplemental Fig. 4). Both TAK1 inhibitors effectively blocked TLR2- and TLR8-induced JNK and p38 phosphorylation, in contrast to the IKKβ inhibitors. IKKβ inhibitors still blocked the phosphorylation of p105 and ERK1/2. This is in agreement with the canonical model of MyD88 signaling where JNK and p38 are downstream of TAK1, whereas activation of p105 and ERK1/2 are controlled by IKKβ (33). The degradation of IkBα was not blocked by any of these inhibitors (Supplemental Fig. 4), and IKKα can probably phosphorylate IkBα, leading to its degradation once IKKβ is lost (34). Failure of TAK1 inhibitors to rescue IkBα from degradation fits with the TAK1-independent nuclear accumulation of p65 (Fig. 7E, 7F), whereas p65 activation may require phosphorylation by IKKβ in addition to degradation of IkBα (33). The TAK1 inhibitor 5Z-7-oxozeaenol efficiently antagonized IKKα/β phosphorylation and IKKβ activity (Supplemental Fig. 4B), whereas NG25 was less efficient (Supplemental Fig. 4A).

We further used gene silencing to examine the role of TLR7, TLR8, IRF5, IKKβ, STING, and TBK1 for IRF5 nuclear accumulation in MDMs upon infection with S. aureus Δlgt (Fig. 7G). Silencing shows that also in MDMs, S. aureus activates IRF5 via TLR8 and IKKβ, confirming the specificity of IRF5 staining and the inhibitor data, whereas TLR7, TBK1, STING, and p65 silencing did not influence IRF5 activation. In contrast, p65 nuclear accumulation was reduced solely with siRNA for p65, verifying the specificity of the nuclear translocation assay also for this factor (Fig. 7H).

We conclude that both primary monocytes and MDMs sense S. aureus via TLR8, and that TLR8 signaling includes a novel TAK1–IKKβ–IRF5 pathway that is required for IFN-β induction and that is blocked by TLR2 signaling. We thus propose a model for TLR8 and TLR2 signaling in monocytes that includes two distinct pathways (Fig. 8).

In this study, we provide evidence for a novel role of TLR8 in sensing of S. aureus by human primary phagocytes. The physiological role of TLR8 is demonstrated by its contribution to the cytokine response induced by the whole bacteria during infection. This is consistent with the findings by Eigenbrod et al. (35) showing a central role of TLR8 in recognition of live Streptococcus pyogenes in human MDMs. Although TLR2 and TLR8 display considerable redundancy in signaling and both contributed to S. aureus–induced production of proinflammatory cytokines such as IL-1α/β, IL-18, and TNF, we show a specific role of TLR8 for the induction of IFN-β and IL-12 via IRF5 activation. TLR8 and IRF5 also contributed to TNF production. It is possible that IFNR signaling could have influenced the TNF response, but this seems less likely given the short incubation time (3 h) used.

The function of human TLR8 as a sensor of bacterial RNA appears similar to murine TLR13 (13, 14). However, whereas TLR13 detects a short sequence of bacterial 23S RNA with high specificity (14, 36), TLR7 and TLR8 have weak sequence specificity and generally detect U-rich RNA (30). Recently it was found that the natural TLR8 ligands are degradation products of U-rich RNA in the form of uridine and short U-containing oligomers that work synergistically (37). In agreement with this model, the commercial TLR13 ligand Sa19, which has a single U residue and is stabilized by thioester backbone, did not activate human monocytes or TLR8-expressing HEK293 cells (not shown). Thus, although human TLR8 and murine TLR13 may seem functionally analogous, their ligand specificities and mechanisms of activation are different. This is also in agreement with the findings of Eigenbrod et al. (35).

The role of IFN-β in S. aureus pathogenesis is controversial, and two recent studies found contradictory effects of IFN-β on the outcome of experimental murine infection (19, 20). In humans, S. aureus forms abscesses with a high local bacterial load that can leak into the surrounding tissue and the bloodstream. The maximum levels of IFN-β produced in response to HK S. aureus exceeded that of HK E. coli and LPS. A more potent induction of IFN-β by HK S. aureus from the exponential phase than the stationary phase may be related to the change in cell wall thickness (38), or it could reflect changes in the amount of stimulatory RNA. Only a fraction (e.g., 20–30%) of the monocyte population stained positive for IRF5 and IRF3 nuclear accumulation after TLR8 or TLR4 activation, respectively. This fits with a stochastic model of IFN-β production in a cell population, which is explained by variations in limiting components at the cellular level (39, 40). TLR8 is also a potent inducer of IL-12 in human monocytes (28, 31), whereas IRF5 regulates macrophage polarization and drives IL-12 and IL-23 production and Th1–Th17 activation (41). Th17 cells may also be important for protection from S. aureus infections (42). We found that lipoproteins of S. aureus, either soluble or in the bacterial cell wall, activated TLR2 and strongly suppressed TLR8-induced production of IFN-β and IL-12 by bacteria being degraded in phagosomes. This may represent a control mechanism to limit IFN-β and IL-12 production induced by extracellular bacteria. It is thus possible that dysregulation of TLR2/TLR8 signaling can lead to disease progression, and it could represent a new immune-evasion target for bacteria such as S. aureus.

Human blood monocytes also sense Borrelia burgdorferi RNA via TLR8, resulting in IFN-β production (43, 44). Although this finding is in agreement with our present study on S. aureus, the proposed signaling mechanisms are different, as they suggested IRF7 to be the central transcription factor involved. We show that TLR8-induced IFN-β production in monocytes and MDMs is dependent on IRF5. The activation of IRF5 by S. aureus and TLR8 ligands was rapid and did not occur in bystander cells not infected by S. aureus, excluding the possibility that IRF5 is triggered via a secondary mediator. Mycobacterium tuberculosis induces IFN-β in mouse macrophages through a nucleotide-binding oligomerization domain–containing protein 2 (NOD2)–receptor interacting protein 2 (RIP2)–TBK1–IRF5 pathway (45). Also, a NOD2–IRF5 mechanism of S. aureus–mediated IFN-β induction in mouse BMDCs was recently reported (46), and IRF5 was activated by TBK1/IKKi and RIP2 in overexpression studies (47, 48). However, we found no induction of IFN-β in RNA-depleted S. aureus lysates, which most likely contain significant amounts of peptidoglycan and bacterial cell wall fragments. Moreover, silencing of TBK1 in MDMs did not affect IRF5 translocation in our studies. Thus, neither NOD2 nor TBK1 seems to be involved in S. aureus–mediated IRF5 activation, arguing against a NOD2–RIP2–TBK1 pathway for S. aureus–induced IFN-β in primary human phagocytes.

In model systems, TLR7 and TLR8 can activate both IRF5 and IRF7, but not IRF3, and in THP-1 monocytic cells TLR7-induced IFN-α is IRF5-dependent (48). In human plasmacytoid dendritic cells, IRF5 rather than IRF7 regulates TLR9-induced IFN-β production together with NF-κB p50 (49). It thus appears that TLR7, TLR8, and TLR9 signaling can involve IRF5 in different cell types. IFN-type I induction by TLR7 and TLR9 in the human plasmacytoid dendritic cell line Gen2.2 is dependent on TAK1 and IKKβ, but independent of IRF7 (34), which is similar to our findings of TLR8-induced IFN-β in primary phagocytes. An essential function of IRF5 for IFN-β induction by TLR7 ligands in Gen2.2 cells was recently demonstrated by Cohen and colleagues (50). They found that IKKβ catalyzes IRF5 phosphorylation, leading to its dimerization and nuclear translocation and resulting in induction of IFN-β and IL-12, but not IL-6. Our finding on TLR8-induced signaling in human primary monocytes and MDMs is similar to this newly described TLR7–TAK1–IKKβ–IRF5–IFN-β pathway in Gen2.2 cells. The function of IKKβ as a kinase activating IRF5 was confirmed by another study (51). Our study using human primary monocytes and MDMs stimulated with live S. aureus captures a more physiological relevant function of this recently identified IKKβ–IRF5 link. Additionally, we demonstrate a regulatory role of TLR2 in this pathway in primary cells.

The function of TAK1 is cell type–dependent (52) and was not necessary for p65/RelA nuclear accumulation in monocytes in our study. A TAK1-independent but MEKK3-dependent pathway of NF-κB activation has been described in mouse and human model systems (53, 54) and could possibly be involved also in human primary monocytes.

Monocytes have high mRNA levels of TLR2, TLR4, and TLR8 and low levels of TLR3, TLR7, and TLR9 (29), which is generally in agreement with our data. Still, the TLR7-specific ligand (R837) induced IFN-β production in monocytes, suggesting that TLR7 is expressed at a functional level. The differential abilities of TLR8 and TLR2 to activate IRF5 via MyD88 might be related to their predominant localization within different cellular compartments, that is, endosomal versus cell surface. TLR2 did not inhibit IFN-β induction via TLR4, TLR7, or cytosolic poly(I:C), which indicate a specificity for TLR8 signaling by TLR2 suppression. It is unclear whether the TLR2-induced feedback mechanism targets IRF5 directly, or whether it acts upstream and interferes more generally with TLR8 signaling. Distinct signaling pathways activated by TLR2 that inhibit TLR8 signaling are a subject of future studies.

In conclusion, we have identified TLR8 as a physiological significant sensor of entire S. aureus and described a novel TLR8–IRF5 signaling axis triggering IFN-β production in primary human monocytes and macrophages antagonized by TLR2. This mechanism may be important for the sensing of infection with S. aureus and possibly other pyogenic bacteria, thus providing new possible targets for pharmacological immunomodulation in conditions such as Gram-positive sepsis.

The imaging was performed at the Cellular and Molecular Imaging Core Facility, Norwegian University of Science and Technology. We thank Dionne Klein and Kjartan Egeberg at the Cellular and Molecular Imaging Core Facility for technical assistance with imaging and Dorte Christansen for help with flow cytometry.

The authors have no financial conflicts of interest.