Concentrations of IFNβ were determined with the VeriKine-HS human IFNβ serum ELISA kit (PBL Assay Science, Piscataway, NJ, USA). TNF and IL-6 ELISA duo-kits were from R&D Systems, and BioPlex assays were from Bio-Rad. The cytokine levels were determined as per the manufacturer’s instructions. The PRR ligands FSL1, CL75, LPS O111:B4, poly(deoxyadenylic-deoxythymidylic) acid (poly-dA:dT; B-DNA), polyinosinic-polycytidylic acid [poly(I:C)], and polyuridylic acid (pU) were purchased from Inviviogen. Poly-l-arginine (pLA) was from Sigma. Lipofectamine 2000 (L2K) was purchased from Invitrogen.

Staphylococcus aureus 113 strain and its isogenc lgt mutant strain (Δlgt) were generously provided by prof. Friedrich Göetz (University of Tübingen, Tübingen, Germany), while the Cowan strain was generously provided by Prof. Timothy Foster (Trinity College, Dublin, Ireland). The GBS NEM316 wild type (wt) and its isogenic lgt mutant strain (Δlgt) were generously provided by prof. Philipp Henneke (University of Freiburg, Germany). The GBS NEM316 β-hemolysin/Christie Atkins Munc-Petersen (CAMP)-deficient strains (ΔcylE, Δcfb, and double deficient ΔcylEΔcfb) have previously been described (26). The E. coli Seattle 1946-strain was purchased from the American Type Culture Collection (ATCC 25922).

E. coli and GBS were grown on tryptic soy agar while S. aureus strains were grown on mannitol salt agar. To prepare bacteria for use in infection experiments, colonies were picked and grown in 5 ml tryptic soy broth (E. coli and S. aureus) or Todd-Hewitt broth (GBS) during vigorously shaking at 37°C overnight (12–18 h). HI was done at 70°C for 30 min. To quantify the bacteria, we initially determined the concentrations of GBS, S. aureus, and E. coli in cultures (stationary phase) by manual microscopy counting, by using a Neubauer Improved counting chamber with 0.02 mm depth (Assistant, Germany) and a Nikon Eclipse E100 phase-contrast microscope with Plan 100× oil objective (NA 1.25). We subsequently determined the OD600 of the cultures using a cell density meter (Ultrospec 10, Amersham Biosciences) for multiple dilutions of each culture within an OD600 range of 0.05–1.20. Standard curves for OD600 versus the corresponding bacteria concentrations were made, and were close to linear in the examined range. An OD600 of 1.00 on the specified equipment correspond to 5.6 × 10⁸/ml of E. coli and 1.5 × 10⁹/ml of GBS and S. aureus. In subsequent experiments the OD600 of the bacterial cultures were used to calculate the bacteria concentration using the standard curves. To calculate the MOI, the bacterial concentrations were converted to CFU, and blood agar counting revealed bacteria-to-CFU ratio of 1:1 for E. coli, 5:1 for S. aureus, and 10:1 for GBS, which reflects that S. aureus and GBS grow in clumps and strings, respectively. The method of quantification was validated for bacterial cultures in both stationary and exponential phases of growth.

Bacterial RNA extraction was conducted in accordance with the procedure described (15) with some modifications. Briefly, bacterial cell extracts were obtained by vortexing both live and HI GBS (grown in the stationary phase) on ice in the presence of glass beads (0.5 mm, Precyllys). Total RNA was purified by the RNeasy Mini kit using DNAse (Qiagen), according to the manufacturer’s protocol. The RNA integrity was examined using Agilent 2100 bioanalyzer (Agilent Technologies, Inc.), and the concentrations were determined with Qbit Fluorometer and Nanodrop (Thermo Fisher Scientific).

Human PBMCs were isolated from buffycoats using Lymphoprep (Axis-Shield) as described by the manufacturer. Monocytes were purified by adherence in culture plates or by negative selection using magnetic beads (Miltenyi Biotec Pan Monocyte Isolation Kit). Monocytes were maintained in RPMI 1640 (Life Technologies) supplemented with 10% pooled human serum.

For infection of monocytes and macrophages, extracellular bacteria were killed after 1 h by addition of gentamicin to 100 μg/ml final concentration.

Monocyte-derived macrophages were generated by differentiation of monocytes in RPMI 1640 with 300 μl 30% pooled human serum for 5–6 days in 96-Well Glass Bottom Microwell Plates (MGB096-1-2-LG-L, Matriplate) or μ-Plate 96 Well ibiTreat (Ibidi). Medium was replaced with RPMI 1640 containing 10% human serum before infection, stimulation, or siRNA treatment.

TLR8 knockout THP1 monocytes (19) were generously provided by prof. Veit Hornung (University of Bonn, Germany). A fluorescent TLR8 construct TLR8-mNeonGreen was made by Gateway cloning. hTLR8 isoform 1 and mNeonGreen (for fluorescent construct) were amplified by PCR from pUno-hTLR8a (Invivogen) and pNCS-mNeonGreen (Allele Biotech), respectively. For non-fluorescent constructs the following PCR primers were used.

TLR8-fw: GGGGACAAGTTTGTACAAAAAAGCAGGCTT CATGGAAAACATGTTCCTTCAG, TLR8-rw: GGGGACCACT TTGTACAAGAAAGCTGGGTTTTAGTATTGC TTAATGGAATCG. For fluorescent construct, the fragments were BP recombined with pDONR221 L1-R5 and L5-L2 vectors to generate multisite pEntry clones. A lentiviral expression clone was generated by LR recombination between pEntry-TLR8-L1-R5, pEntry-mNeonGreen-L5-L2 and pLenti-CMV-Puro-DEST (w118-1) (a gift from Eric Campeau, Addgene plasmid # 17452), giving a C-terminal in-frame fusion of TLR8 to mNeonGreen. For non-fluorescent TLR8 expressing construct, fragment was BP recombined with pDONR221 vector followed by LR recombination with pLenti-CMV-Puro-DEST (w118-1) vector. All constructs were verified by Sanger sequencing (GATC biotech). Lentiviral particles were produced in HEK293T cells using a third generation lentiviral packaging system (Addgene plasmids # 12253, 12251, and 12259) and concentrated using PEG precipitation as described elsewhere (27). For fluorescent construct, TLR8 knockout THP1 cells were spin-transduced in the presence of 8 µg/mL polybrene and 1 mg/ml Synperonic-F108 (Sigma) (28), selected on 0.6 µg/ml puromycin and subjected to one round of FACS to select for a moderate expression level. For non-fluorescent TLR8 overexpression, THP1 WT cells (ATCC) were transduced by viral particles containing a pLenti-CMV-Puro-DEST-TLR8 construct, or with virus containing empty vector (for control cell line). Cells were grown for selection on 1 µg/ml puromycin for 4 weeks. THP1 cells were differentiated using 100 ng/ml phorbol 12-myristate 13-acetate [PMA (Sigma)] for 3 days, followed by 1 day rest in fresh media before infection.

Cells were lysed in lysis buffer and denatured in 1× NuPage LDS (Invitrogen) sample buffer supplemented with 25 mM DTT for 10 min at 70°C. The samples were separated on 4–12% Bis-Tris polyacrylamide gel and transferred to a nitrocellulose membrane using the iBlot Dry Blotting System (Invitrogen). The blots were incubated with horseradish peroxidase-conjugated immunoglobulin’s (DAKO) and developed with Super Signal West Femto Maximum Sensitivity Substrate (Thermo Scientific) before images were obtained with Odyssey Fc Imaging System (LI-COR). TLR8 overexpression in THP-1 subclones was confirmed by Western blotting using rabbit antihuman TLR8 XP mAb [Cell Signaling Technology (CST), no. D3Z6J]. Activation of IRF3 was done using rabbit antiphospho-IRF3 mAb (CST, no. D6O1M). Anti GAPDH (Abcam, #ab8245) served as a loading control.

A pool of four individual ON-TARGETplus siRNAs (Dharmacon) was transiently transfected using siLentFect (Bio-Rad), yielding a final concentration of 5 nM siRNA. The transfection was repeated after 3 days, and the silenced MDM were infected with bacteria or stimulated with ligands for 4 h. RNA was isolated with an RNeasy 96 Plus kit (Qiagen), cDNA was transcribed with a Maxima cDNA synthesis kit (Thermo Fisher Scientific), and relative quantification by qPCR was done with StepOnePlus using TaqMan probes (Life Technologies) and Perfecta qPCR FastMix from Quanta. Probes used were: IFNβ, Hs01077958_s1; TNF, Hs00174128_m1; IL-6 Hs00985639_m1; IL-12A Hs1073447_m1; TBP, Hs00427620_m1, IKKβ Hs00233287_m1, cGAS/MB21D1 Hs00403553_m1, MyD88 Hs00182082_m1, STING/TMEM173 Hs00736958_m1, TLR7 Hs00152971_m1, TLR8 Hs00607866_mH, IRF5 Hs00158114_m1, and TBK1 Hs00179410_m1. TBP served as endogenous control, and relative expression was calculated as fold induction by stimulation or infection.

Immunofluorescence staining of TLR8 and transcription factors and quantification of nuclear accumulation by high content screening (Scan^R system, Olympus) was done as previously described (24). The following antibodies were used: mouse antihuman IRF5 mAb (Abcam, #10T1), rabbit antihuman IRF3 XP mAb [Cell Signaling Technology (CST), no. 11904], rabbit antihuman p65/RelA XP mAb (CST, no. 8242), rabbit antihuman p65 A (Santa Cruz Biotechnology, no. sc-109), rabbit antihuman IRF1 XP mAb (CST, no. 8478), and rabbit antihuman TLR8 XP mAb (CST, no. D3Z6J). For TLR4 IF staining monocytes were washed by ice-cold PBS twice, fixed by cold methanol-acetone (1:1) at −20°C overnight, rehydrated in PBS for 1 h, and blocked by 20% human serum in PBS for 30 min. After blocking, cells were incubated with primary antibodies diluted in 2% human serum [2 µg/ml rabbit anti-TLR4 IgG (H-80) or normal rabbit IgG (Santa Cruz Biotechnology)] at +4°C overnight, washed three times by PBS with 2% human serum, incubated with secondary A488-labeled goat antirabbit IgG (Life Technologies). After three washes, cells were left in PBS at +4°C prior to confocal microscopy imaging, which was done using Leica SP8 with a 63× objective (NA 1.4).

Statistical analyses were done on data merged from independent experiments (not replicates). The number of independent experiments (n) is given in the figure legends. Prior to statistical analyses, the data were log transformed to increase the likelihood of a Gaussian distribution. Significant differences was tested by one-way RM ANOVA with Dunette’s posttest, or by two-way RM ANOVA with Bonferroni multiple comparison posttest. Significance levels: *p < 0.05, **p < 0.01, and ***p < 0.001. All analyses were performed with GraphPad Prism v5.03.

The research was carried out according to the standard institutional safety procedures (biosafety level II) at the Norwegian University of Science and Technology.

We recently showed how IFNβ and IL-12-p70 production by human primary monocytes upon S. aureus infection is mediated by TLR8-IRF5 signaling (24). Significant reduction in TLR8-IRF5 activation is seen with S. aureus strains activating TLR2 and is also characteristic of S. aureus when it is grown to the stationary phase compared to the exponential phase (24). To investigate the role of TLR2 and growth phases on GBS induced cytokines, we infected primary human monocytes with viable GBS NEM316 strains (GBS wt) and its isogenic lipoprotein deficient strain (GBS Δlgt) which does not activate TLR2 (7). S. aureus wt and S. aureus Δlgt strains were included for comparison. With bacteria from the stationary growth phase both GBS wt and GBS Δlgt induced more IFNβ and TNF than S. aureus wt strain (Figures 1A,B), while induction of IL-6 was not significantly different between the species (Figure 1C). There were no significant differences in cytokine levels induction by GBS wt and GBS Δlgt strains, suggesting that GBS produced little TLR2-activating lipoproteins, or that the mechanism of cytokine induction by GBS was not affected by TLR2 costimulatory molecules. The low IFNβ and TNF induction by S. aureus wt and S. aureus Δlgt compared to GBS was a characteristic of the stationary growth phase, as similar cytokine induction was for both species in the exponential growth phases (Figure S1 in Supplementary Material). The level of IFNβ production was in particular suppressed for S. aureus wt strain in stationary phase and when the bacterial culture media was included, suggesting both TLR2-mediated suppression of TLR8-IRF5 inhibition and growth-phase dependent changes of S. aureus, in agreement with our previous findings of Ref. (24). In contrast, GBS wt and GBS Δlgt induced similar cytokine levels in either condition (Figure S1 in Supplementary Material). Infection of MDM showed similar differences between GBS and S. aureus for the induction of IFNβ, although the level of cytokine induction was lower (Figure 1D). Altogether, GBS induce more IFNβ and TNF compared to S. aureus in stationary growth phase, and GBS-mediated cytokine induction is mainly independent of TLR2 agonistic GBS lipoproteins.

IFNβ induction by GBS in murine macrophages and human THP-1 cells is dependent on β-hemolytic activity, giving phagosomal damage and leakage of GBS DNA into the cytosol (26, 29). To examine the possible role of GBS hemolytic activity for cytokine induction in primary cells, we compared the GBS wt and GBS Δlgt strains to isogenic strains deficient in β-hemolysin (ΔcylE) and cohemolysin/CAMP factor (Δcfb), as both these factors can contribute to phagosomal pore formation and the induction of IFNβ (26). The hemolysin-deficient phenotype was verified on blood agar (not shown). Infection of primary monocytes revealed that neither IFNβ nor TNF production was dependent on hemolysin (Figure 2A). On the contrary, at the transcriptional level, hemolysin deficiency slightly increased the IFNβ and TNF induction, suggesting that hemolysins to some extent can attenuate the cytokine responses (Figure 2B). The reasons for the tendency of hemolysin mutant strains to induce slightly higher cytokine transcript levels relative to wt bacteria are not entirely clear, but may be related to subtle hemolysin dependent cytotoxic effects.

To examine the effect of TLR2 signaling on cytokine induction by GBS, human primary monocytes were infected with viable GBS and with or without costimulation with the TLR2 ligand FSL-1. FSL-1 costimulation strongly reduced the production of IFNβ by both the GBS wt strain (Figure 2C) and the GBS Δlgt (Figure S2A in Supplementary Material). Analysis of additional cytokines induced by the GBS wt strain revealed a similar effect of FSL-1 costimulation on IL-12-p70 production (Figure 2C), while FLS-1 had no effect the induction of TNF, IL-1β, and MCP-1, but elevated the levels of IL-6 and IL-10 (Figures S2B–F in Supplementary Material). We previously showed that the uptake of S. aureus in monocytes increases by FSL-1 costimulation, while the attenuation of IFNβ and IL-12 production is a result of cross-talk with TLR8-IRF5 signaling, and not a direct effect on the bacteria (24). The similar effects of FSL-1 costimulation on GBS-induced cytokine production indicate that sensing of also this bacterium might depend on endosomal TLR8-IRF5 signaling.

We infected monocytes with stationary grown S. aureus, GBS and E. coli and compared the nuclear accumulation of IRF5 and NF-kB subunit p65 (RelA). Nuclear accumulation of IRF5 was dependent on the bacterial dose and was considerably stronger with GBS compared to E. coli and S. aureus (Figure 3A). In contrast nuclear accumulation of p65 was induced to a similar extent by GBS and E. coli infection, but weaker by S. aureus (Figure 3B). The cells responding to GBS infection were mainly double-positive for IRF5 and p65 nuclear staining, while with E. coli only some cells showed nuclear IRF5 staining (Figure 3C). The differential activation of IRF5 with GBS and S. aureus correlated with the induction of IFNβ with these bacteria (Figure 1A), which suggests that TLR8-IRF5 signaling may be more important for sensing of GBS than stationary grown S. aureus. GBS infection did not trigger IRF3 phosphorylation, while this was clearly seen after LPS or E. coli stimulation (Figure S3 in Supplementary Material). This reflects the important role of IRF3 in TLR4-mediated induction of IFNβ from the endosome via the TRAM-TRIF-TBK1-IRF3 signaling pathway (30).

To clarify the roles of IRF5 and IRF3 in IFNβ production by monocytes induced by distinct PRR ligands, monocytes were stimulated with the IFNβ inducing PRR ligands CL75, LPS, pdA:dT (dsDNA), and poly(I:C) (dsRNA). The nucleic acid ligands were premixed with Lipofectamine 2000 (L2K) for cytosolic transfection of the cargo, in contrast to premixing with cationic delivery agents such as pLA which gives endosomal localization (31). After 30, 60, and 120 min of treatment the cells were fixed and immunostained for IRF3 and IRF5 nuclear accumulation. Cytosolic transfection with pdA:dT gave a weak but significant IRF5 relocalization upon prolonged incubation, while no effect was seen with poly(I:C) transfection (Figure 4A). IRF5 was rapidly activated by the TLR8-ligand, but not at all by the TLR4-ligand (Figure 4A), in agreement with our previous study (24). Thus, in addition to the strong and rapid IRF5 relocalization by TLR8 signaling, cytosolic dsDNA sensing mechanisms can to some extent activate IRF5 nuclear translocation in human primary monocytes. In addition to LPS, TBK1-IRF3 activation is mediated via the cytosolic dsRNA sensing helicases melanoma differentiation-associated gene 5 (MDA5) and retinoic acid-inducible gene-I (RIG-I), and also the dsDNA sensors RNA polymerase III, DEAD box polypeptide 41 (DDX41), IFN inducible protein 16 (IFI16), DNA-dependent activator of IFN regulatory factors (DAI), and cyclic GMP-AMP synthase (cGAS) (32). Thus as expected both LPS and transfection of dsDNA and dsRNA strongly activated IRF3 translocation, while the TLR8 ligand did not. Moreover, IRF1 was not clearly activated by any of the ligands (Figure S4 in Supplementary Material), and neither was IRF7 activation seen (24). However we can not exclude the possibility that also other IRFs might be involved. We further examined the effect of TLR2-costimulation on IRF5 activation by the TLR8 ligand polyuridylic acid (pU) in complex with pLA, and IRF5 and IRF3 activation by cytosolic dsDNA (pdA:dT). While FSL-1 impaired IRF5 activation with pU/pLA, as well as pU/pLA mediated induction of both IFNβ (24) and IFNα (not shown), it influenced neither IRF5 nor IRF3 activation by cytosolic pdA:dT (Figure 4B). IFNβ induction by a TLR7 ligand in monocytes is not affected by FLS-1 costimulation (24), and the same holds true for different TLR7 ligands in human full blood, while a TLR9 ligand does not stimulate cells in whole blood or monocyte cultures (not shown). The TLR2-antagonistic effect thus seems to be restricted to TLR8-IRF5 activation in these model systems and does not inhibit IRF5 activation via other signaling pathways.

We further investigated the effect of TLR2 costimulation and HI on IRF5 and p65 nuclear translocation induced by GBS and E. coli. Human primary monocytes were infected with viable bacteria or stimulated with HI GBS, HI E. coli or the TLR8 ligand CL75, and nuclear staining was subsequently performed. No significant activation of IRF5 was seen with HI E. coli, while HI GBS activated IRF5 at a reduced level compared to viable GBS. The viable bacteria also activated p65 more potently than HI bacteria (Figure 5). The reduced potency of the HI GBS correlated with loss of RNA integrity (Figure S5 in Supplementary Material), suggesting that RNA is a vita-PAMP of GBS. This finding is similar to what has been described for E. coli, where inactivation by either heat or antibiotics resulted in rapid RNA degradation (33). Interestingly, while activation of IRF5 by TLR8 and GBS was reduced by TLR2 costimulation, IRF5 activation by E. coli was not affected by TLR2 ligand treatment. This suggests that E. coli activates IRF5 by a mechanism that is independent of TLR8 and possibly dependent on cytosolic nucleic acid sensing.

Several lines of evidence suggest that GBS is sensed by TLR8 in human primary monocytes. To examine whether TLR8 is recruited to GBS phagosomes, we overexpressed TLR8 in THP-1 cells. Using TLR8 knockout cells (19), we overexpressed a TLR8 mNeonGreen construct (TLR8mNG). The cells were differentiated to macrophages with PMA and infected with viable GBS. Imaging of fixed cells revealed accumulation of TLR8 on GBS phagosomes in roughly one half of the cells with medium level of fluorescent TLR8 expression (Figure 6A). We also examined TLR8 localization in primary monocytes by IF. The specificity of TLR8 immunostaining was tested using THP-1 cells overexpressing full length TLR8 (Figures S6A,B in Supplementary Material). Staining of primary monocytes infected with viable GBS revealed accumulation of endogenous TLR8 on phagosomes (Figure 6B), similar to the findings with the TLR8-mNG recruitment in THP-1. We also examined recruitment of TLR9 in TLR9-mNG expressing THP-1 cells, but TLR9 was mainly localized to the endoplasmatic reticulum and did not accumulate on GBS (not shown). IF of TLR4 in primary monocytes did not reveal any TLR4 recruitment to S. aureus or GBS phagosomes, while E. coli phagosomes were clearly stained (Figure S6C in Supplementary Material). These results are in agreement with our previous study using HI bacteria (30).

To determine the role of TLR7, TLR8, IRF5, and STING (stimulator of interferon genes) for cytokine induction in MDM upon infection with viable GBS and E. coli, we performed siRNA-mediated gene silencing, which was efficient for all targets examined (Figure S7A in Supplementary Material) The pU/pLA induced cytokines were strongly reduced after TLR8 silencing (Figure 7). TLR7 knockdown did not significantly affect cytokine induction by either TLR8 ligand or bacteria. In contrast, silencing of TLR8 and IRF5 strongly impaired the GBS activated induction of IFNβ and IL-12-p35, and to some degree also TNF (Figure 7). IL-6 induction by GBS was not significantly affected by TLR8 or IRF5 knockdown, suggesting that TLR8 is redundant with other PRRs for the induction of this particular cytokine. Silencing IRF5 also affected pU-mediated induction of IL-6 less efficiently compared to the other cytokines and suggests that IL-6 transcription is mainly dependent on the more general factors NF-kB and MAPKs. None of the GBS induced cytokines was STING dependent (Figure 7). Induction of cytokines by E. coli was independent of TLR8, but maximal IFNβ and IL-6 induction was partly dependent on IRF5. A TLR8-independent mechanism of IRF5 activation by viable E. coli is supported by the nuclear translocation studies, because this was insensitive to TLR2 costimulation (Figure 5). Moreover, a role of STING for the induction of IFNβ by E. coli is indicated, suggesting the possible involvement of a STING-IRF5 signaling axis (Figure 7). STING can be activated by cyclic dinucleotides produced by cGAS upon cytosolic DNA sensing, or by cyclic dinucleotides produced by the bacteria (34). Silencing of cGAS in MDM did not attenuate E. coli-mediated induction of IFNβ (Figure S7B in Supplementary Material), suggesting that STING is not activated by viable E. coli via the DNA-cGAS-cGAMP (cyclic-di-GMP-AMP) pathway. MyD88 or IKKβ knockdown also did not inhibit IFNβ induction by E. coli (Figure S7B in Supplementary Material), which supports that endosomal TLR signaling is not involved (Figure 7). TBK1 silencing, on the other hand, attenuated the E. coli-mediated IFNβ production (Figure S7B in Supplementary Material), which is expected given its central role in the TLR4 signaling pathway upstream of IRF3 (30).

In conclusion, TLR8-IRF5 signaling is a major contributor to the sensing of GBS in primary human monocytes and MDM by inducing IFNβ, IL-12, and TNF. In comparison with GBS, TLR8 is less strongly activated by stationary grown S. aureus and does not contribute to cytokine induction by viable or HI E. coli. Still, IRF5 plays a role in cytokine induction by viable E. coli.