Macrophages Are a Potent Source of Streptococcus-Induced IFN-β

In tissues immediately adjacent to mucocutaneous surfaces, which are colonized by commensal bacteria, mononuclear phagocytes (e.g., macrophages [MΦ] and dendritic cells [DC]) form the quantitatively largest pool of immune cells. MΦ in the dermis (skin) and the lamina propria (gastrointestinal tract) exceed most other tissues by a high turnover with bone marrow–derived monocytes. It seems plausible that the cellular input and the reception of epigenetic marks under the influence of the resident microflora propagate diversity in immune cells at these sites (1–3). For example, circulating monocytes replenish the MΦ pool under homeostatic conditions, whereas they give rise to DC during colitis, emphasizing niche- and demand-driven plasticity of these phagocytes (4). Yet the plasticity of DC and MΦ driving the acquisition of tissue specific identities remains incompletely defined (2).

In vitro cultures of bone marrow progenitors with the growth factors M-CSF and GM-CSF have been instrumental for the understanding of DC and MΦ differentiation. The M-CSF culture has the advantage of yielding a relatively pure MΦ population and is most commonly used for studying these cells. In contrast, GM-CSF promotes the generation of a mixed DC and MΦ culture and is employed more frequently in the DC field (5, 6). M-CSF and GM-CSF are produced by a broad variety of immune and nonimmune cells. Next to their eponymous function as growth factors for hematopoietic progenitor cells, they also modulate survival, activation, or differentiation of myeloid cells. M-CSF induces proliferation of MΦ and monocytes without major alterations of their activation status, whereas GM-CSF transcriptionally activates MΦ and DC. In contrast to M-CSF, circulating GM-CSF is hardly detectable under homeostatic conditions (7), but its concentration rapidly rises during tissue infections and damage (8). Accordingly, GM-CSF has been targeted to modulate inflammation, such as in IL-23–driven colitis (9).

Streptococcus agalactiae, or group B Streptococcus (GBS), is a common colonizer of the human genital and gastrointestinal tracts yet causes life-threatening infections in neonates, pregnant females, and elderly adults (10–12). Thus, GBS is a model pathobiont for the understanding of mucocutaneous antibacterial resistance. It appears that the development of cellular innate immunity in general, as well as the type I and II IFN and the inflammatory cytokine responses, are essential for host resistance against GBS in vivo (13, 14). Cytokine formation in monocytes and MΦ relies on the recognition of GBS lipoproteins via TLR2 and TLR6, and bacterial ssRNA by endosomal TLRs 8 (in humans) and 13 (in mice) (15–18). In contrast, conventional DC (cDC) were reported to be the predominant source of type I IFN in GM-CSF–supplemented bone marrow cultures (GM cultures) stimulated by GBS, yet definition of cDC solely relied on the expression of CD11c in this work. The IFN response was shown to rely on TLR7 engagement driving the transcription factor IRF1 (13). Other work showed type I IFN response in MΦ to depend on cytosolic sensing of GBS nucleic acids by cGAS (19).

The recent gain in understanding of MΦ heterogeneity and adaptability in the tissue prompted us to revisit cell lineage and TLR-dependent induction of type I IFN by GBS. We embarked on precisely monitoring the phenotypic dynamics in mixed mouse MΦ/DC cultures with GM-CSF, which requires exact characterization of cellular identities. We found that bone marrow cultured in GM-CSF differentiated into four mononuclear phagocyte populations. Among these, only the two MΦ-like subsets, and not the DC-like subsets, used TLR7 for IFN-β formation in response to GBS. Moreover, subsequent to the encounter with streptococci, MΦ changed their immunophenotype and morphology and picked up some DC markers and a clustered behavior. Thus, differentiated MΦ form type I IFN and modulate their identity by incorporating DC-like features in response to commensal bacteria.

All mice were on C57BL/6J or C57BL/6N genetic background. Mice harboring an UNC93B1-H412R (3D) mutation were kindly provided by M. Freudenberg (University of Freiburg) (16). Ifn-β^wt/Δβ-luc and Ifnar^−/− mice were kindly provided by P. Stäheli (University of Freiburg). OT-II mice were kindly provided by E. Pearce (Max Planck Institute of Immunobiology and Epigenetics, Freiburg, Germany). Tlr7^−/− mice were purchased from The Jackson Laboratory.

Animals used for experiments were between 6 and 12 wk old and age and sex matched.

GBS strains NEM316 or BM110 were grown to midlog phase in Todd Hewitt broth (20). Bacterial concentration was determined with an optical spectrophotometer (OD₆₀₀). CFU were counted by serial dilutions on blood agar plates (Columbia agar + 5% sheep blood; BioMérieux). For heat fixation (fixed GBS), bacteria were incubated for 30 min at 80°C shaking after adjusting the cell number to 10¹⁰ CFU/ml as previously described (16).

Bacteria (GBS strain BM110) were grown to midlog phase, washed with PBS twice, and resuspended in PBS at ∼2 × 10⁶ CFU/ml. A total of 0.5 ml suspension of bacteria or vehicle (PBS) were injected i.p. After 18 h, mice were sacrificed and peritoneal lavage was performed. Cells were sorted by FACS (MoFlo Astrios) and subsequently analyzed by quantitative RT-PCR (qRT-PCR).

Six- to twelve-week-old donor mice were sacrificed and sprayed with 70% ethanol. Then, femurs and tibias were isolated, rinsed with sterile PBS, and opened. Bone marrow cells were flushed with a 27-gauge needle and passed through a 70-μm cell strainer. Pelleted cells were resuspended in RPMI medium supplemented with 10% FBS, ciprofloxacin (10 μg/ml), and either GM-CSF (20 ng/ml; PeproTech) or M-CSF (20 ng/ml; PeproTech) and plated in T75 cell culture flasks (Greiner Cellstar) at 37°C and 5% CO₂.

For GM-CSF–supplemented bone marrow cultures, half of the culture medium was removed at day 2, and new medium containing 40 ng/ml GM-CSF was added. At day 3 and 6 of culture, the medium was discarded and fresh medium containing 20 ng/ml GM-CSF was added. At day 8, nonadherent and adherent cells were pooled by scraping the flasks and collecting the cell suspension.

To obtain M-CSF–derived MΦ, culture medium was changed every 3 d with fresh medium containing 20 ng/ml M-CSF. At day 8 of culture, supernatant was discarded and cells were harvested by scraping the flask and washing gently with PBS.

Six-to twelve-week-old donor mice were sacrificed and sprayed with 70% ethanol. Peritoneal lavage was performed with 5 ml ice-cold sterile PBS. Cells were washed and resuspended in RPMI medium supplemented with 10% FBS and ciprofloxacin (10 μg/ml) and plated in a 96-well flat-bottom plate (Thermo Fisher Scientific), 80,000 cells per well. The next day the cells were washed with fresh medium.

For flow cytometry, cells were washed with staining buffer (2 mM EDTA and 2% FCS in PBS), and 10⁶ cells were stained for 30 min at 4°C in 100 μl volume. Subsequently, cells were washed and analyzed with a 10-color flow cytometer (Gallios; Beckman Coulter) and the Kaluza software (version 1.5a; Beckman Coulter). The following anti-mouse Abs were used: CD11b PE-Cy7 (eBioscience); CD11c FITC (BD Biosciences); CD115 allophycocyanin (Miltenyi Biotec); F4/80 allophycocyanin (Miltenyi Biotec); F4/80 PE (eBioscience); MHC class II (MHC II) eFluor 450 (eBioscience); CD86 FITC (BD Biosciences); CD103 PE (BD Biosciences); CD135 PE (BioLegend); CD64 PerCP/Cy5.5 (BioLegend); CD206 PE (BioLegend); CD45 eFluor 450 (Invitrogen); CCR7 PerCP/Cy5.5 (BioLegend); CD3 PE (eBioscience); Ly6G PE (BD Biosciences); Ki67 PE/Cy7 (eBioscience); CD4 PerCP/Cy5.5 (eBioscience); CD19 AF488 (eBioscience).

For ELISA, 80,000 cells per well were seeded in a 96-well flat-bottom plate (Thermo Fisher Scientific). After resting overnight, cells were stimulated with LPS (100 ng/ml; InvivoGen) or fixed or live GBS. Two hours after the infection with live bacteria, 250 μg/ml gentamicin (Sigma-Aldrich) was added to the culture to stop bacterial growth. After 24 h, IFN-β, TNF-α, and IL-6 in the supernatant were quantified by ELISA kits according to the manufacturer’s instructions.

For TLR7 inhibition, cells were pretreated for 15 min with different doses of INH-ODN 24888 (0, 1, and 10 μM; BioSpring) and afterward stimulated with fixed GBS (10⁷/ml) for 24 h.

For FACS analysis of surface marker expression, cells were sorted by FACS (MoFlo Astrios), resuspended in RPMI medium with 10% FBS plus ciprofloxacin (10 μg/ml), and plated in 96-well plates (from 50,000 to 100,000 cells per well). The next day, cells were stimulated with medium containing the indicated stimulants for 2, 4, 8, and 24 h at 37°C. After stimulation, cells were washed once and subsequently stained and analyzed by FACS.

Cells were sorted by FACS (MoFlo Astrios), resuspended in RPMI medium with 10% FBS plus ciprofloxacin (10 μg/ml), and plated in 24-well plates (200,000 cells per well). The next day, WGA488-labeled, fixed GBS was added (10⁷/ml) for 0, 20, and 60 min. After stimulation, cells were washed three times and subsequently stained and analyzed by FACS.

Cells were sorted by FACS (MoFlo Astrios), resuspended in RPMI medium with 10% FBS plus ciprofloxacin (10 μg/ml), and plated in 48-well plates (from 50,000 to 100,000 cells per well). The next day, cells were stimulated with medium containing the indicated stimulants for 2 or 24 h at 37°C. Total RNA was extracted using the RNeasy Micro kit, according to instruction manual (Qiagen). qRT-PCR was performed as previously described previously (21). For RNA extraction from resting cells as well as for RNA sequencing analysis, cells were directly sorted into RLT lysis buffer. The following mouse primer sequences were used (5′-3′): murine GAPDH fw 5′-ACTCCACTCACGGCAAATTC-3′, murine GAPDH rev 5′-TCTCCATGGTGGTGAAGACA-3′; murine IFN-β fw 5′-TGTCCTCAACTGCTCTCCAC-3′, murine IFN-β rev 5′-CCTGCAACCACCACTCATTC-3′; murine IL-10 fw 5′-CCCTTTGCTATGGTGTCCTT-3′, murine IL-10 rev 5′-TGGTTTCTCTTCCCAAGACC-3′; murine Stat3 fw 5′-GCACCTTGGATTGAGAGTCA-3′, murine Stat3 rev 5′-CCCAAGAGATTATGAAACACCA-3′; murine RelB fw 5′-GTTCCAGTGACCTCTCTTCCC-3′, murine RelB rev 5′-CCAAAGCCGTTCTCCTTAATGTA-3′; murine Maf fw 5′-CTGCCGCTTCAAGAGGGTGCAGC-3′, murine Maf rev 5′-GATCTCCTGCTTGAGGTGGTC-3′; murine TLR2 fw 5′-TTTGCTGGGCTGACTTCTCT-3′, murine TLR2 rev 5′-AAATCTCCAGCAGGAAAGCA-3′, murine Cxcl10 fw 5′-GGATCCCTCTCGCAAGGA-3′, murine Cxcl10 rev 5′-ATCGTGGCAATGATCTCAACA-3′.

Eighty thousand sorted cells were seeded in a 96-well plate, and pictures were taken 24 h after seeding or after stimulation using the Axio Vert.A1 microscope and the Axiocam503 mono (both from ZEISS). Pictures were then analyzed and edited using the ZEN blue software.

Fifty thousand sorted cells were seeded in an eight-well poly-l-lysine–treated μ-slide (ibidi) and stimulated with fixed and WGA488-labeled GBS for 4 h at 37°C when indicated. After stimulation cells were fixed in 4% PFA, washed twice with 0.05% TBST, and subsequently permeabilized with 0.2% TBST. Then, cells were stained with phalloidin and DAPI for 30 min, washed, and finally mounted with mounting medium (Dako) for microscopic analysis with the LSM880. Pictures were then analyzed and edited using the ZEN black software. Cells size was measured in area per cell (square micrometer).

RNA quality and quantity were measured with Qubit RNA BR (Thermo Fisher Scientific) and Bioanalyzer RNA 6000 Nano (Agilent). One hundred nanograms of total RNA was used for the library preparation with the TruSeq Stranded mRNA Library Prep Kit (Illumina) according to the manufacturer’s protocol (TruSeq Stranded mRNA Sample Preparation Guide). Next-generation sequencing (paired end and 100 bp read length, 6 × 10⁷ raw reads per sample) was performed with HiSeq 4000 instrument (Illumina).

Illumina bcl2fastq (version 2.19) was used for demultiplexing of sequenced reads. Adapter trimming was performed with Skewer (version 0.2.2) (22). Trimmed raw reads were aligned to the murine reference genome mm10 using STAR (version 2.5.2b) (23) with default parameters. STAR was also used for read counting to quantify transcripts. Differential expression between the groups was analyzed with the package DESeq2 (24) in R (R Core Team 2015). A p value cutoff of 0.05 was used to identify significantly differentially expressed genes (DEG).

Principal component analysis was performed using rlog-transformed data of the DESeq2 package (24) in R.

Cells were sorted by FACS (MoFlo Astrios), resuspended in RPMI medium with 10% FBS plus ciprofloxacin (10 μg/ml) and either GM-CSF or M-CSF (20 ng/ml; PeproTech), and plated in 24-well plates (from 200,000 to 300,000 cells per well). The next day, cells were plated into U-shaped 96-well plate (40,000 cells per well) and were stimulated with OVA peptide (0.1 μg/ml) for 2 h at 37°C.

CD4 T cells were isolated from spleen using a CD4 isolation kit (MACS Miltenyi Biotec) following manufacturer’s instructions, after which they were stained with Cell Trace Violet (Thermo Fisher Scientific) for 30 min at 37°C. The reaction was quenched by adding FCS.

CD4 T cells were cocultured in the U-shaped 96-well plate (200,000 cells per well) for 72 h and subsequently analyzed by FACS.

GM-CSF–supplemented bone cultures were carried out for 8 d as described and sorted into populations A to D according to their surface marker expression. On day 8, 80,000 cells were seeded into a white 96-well plate and stimulated for 24 h with fixed GBS (10⁷/ml). Cells were washed, and 15 mg/ml luciferin (Sigma-Aldrich) in PBS was added before consequent luminescence measurement using the EnVision Plate Reader (PerkinElmer).

Skin of Ifn-β^wt/Δβ-luc mice was infected as previously described (17). Briefly, bacteria (GBS strain BM110) were grown to midlog phase, washed with PBS twice, and resuspended in PBS at ∼3 × 10⁸ CFU/ml. A total of 10 μl suspension was injected intradermally into the ear pinna of anesthetized mice (i.p. ketamine/xylazine).

After 24 h, the ears were subjected to enzymatic digestion by Dispase (1 mg/ml; STEMCELL Technologies), Collagenase II (2 mg/ml; PAA), and DNase I (0.8 mg/ml; Roche) in HBSS for 2 h at 1400 rpm with shaking at 37°C. After digestion, the samples were filtered with a 40-μm cell strainer. Fifteen thousand to twenty thousand sorted MΦ were plated into a 96-well plate, and 300 μg of luciferin was added. Luminescence was measured immediately on EnVision reader.

Comparative data were analyzed by two-tailed Student t test. Data are presented as mean ± SEM; p values ≤ 0.05 were considered statistically significant. Statistical analysis was performed with GraphPad Prism 7.

Helft et al. (5) recently showed that GM-CSF cultures give rise to MΦ and cDC, both of which express CD11c and MHC II. We revisited and extended this model by carefully comparing surface marker expression of bone marrow cell cultures supplemented with either GM-CSF or M-CSF (both for 8 d). As expected, M-CSF supplementation led to a phenotypically homogenous population, which was defined by F4/80^hi, CD11c^int, CD206^hi, CD11b^hi, CD115^hi, and MHC II ^int cells. In contrast, GM cultures could be separated into CD115^hi and CD115^lo cells, most likely corresponding to MΦ and DC, respectively. We found that GM-CSF–derived MΦ expressed higher levels of CD11c as well as lower levels of F4/80, CD206, and CD11b than M-CSF–derived MΦ, whereas MHC II expression was approximately similar (Fig. 1A). In GM-CSF cultures, DC expressed high levels of CD86 and CD103, were only loosely adherent, and had a rounded morphology with many dendrites. In contrast, MΦ were CD86^lo, CD103^lo, and CD206^int. They were firmly adherent and had a flat and spread morphology with lamellipodia (Fig. 1B, 1C). The phagocytic activity was similar in MΦ and DC (Fig. 1D). In summary, MΦ derived from M-CSF and GM-CSF cultures exhibited profound immunophenotypic and morphological differences. Conventional surface markers like CD11c and MHC II alone were not suitable to discriminate in vitro generated DC from MΦ.

In 2009, Mancuso et al. (13) reported on the importance of TLR7, the adaptor protein MyD88, and the transcription factor IRF1 for GBS recognition and the subsequent type I IFN response in cDC from GM-CSF cultures but not MΦ from M-CSF. In this study, we embarked on better defining the cellular source of GBS-induced IFN-β in GM cultures, which were sorted into CD115^hi MCH II^int MΦ and CD115^lo MHC II^hi DC and stimulated with fixed and live GBS. Surprisingly, only MΦ, but not DC, produced IFN-β and TNF-α in response to all stimuli (Fig. 2A, 2B). In contrast, GBS-stimulated DC produced IL-6 in substantial amounts (Fig. 2C). In accordance with the results of Mancuso et al., M-CSF–generated MΦ did not produce IFN-β upon GBS stimulation (Supplemental Fig. 1A). To determine the role of endosomal TLRs in the IFN-β response, we analyzed UNC93B1-deficient MΦ from GM culture. The response to fixed GBS was completely abrogated in UNC93B1-deficient cells, indicating an essential role of endosomal TLRs under these conditions (Fig. 2D). However, UNC93B1-deficient MΦ responded with the production of IFN-β when stimulated with live GBS with a multiplicity of infection (MOI) equal to 1. This suggested that recognition of live GBS was at least partially independent of endosomal TLRs. Next, we wondered whether the type I IFN response in MΦ from GM cultures was dependent on TLR7. To test this, we stimulated bone marrow cells from Tlr7^−/− mice with fixed GBS. Whereas IFN-β was barely produced in TLR7-deficient cells following stimulation with GBS, TNF-α production remained unaffected (Fig. 2E, 2F). Similar results were obtained when wild-type (wt) cells were pretreated with INH-ODN 24888, a synthetic antagonist of TLR7 (25, 26), before stimulation with fixed GBS. The IFN-β production was decreased by INH-ODN 24888 in a dose-dependent manner, whereas TNF-α production remained unaltered (Supplemental Fig. 1B). These data suggested that only GM-CSF–derived CD115^hi MΦ, and not GM-CSF–derived CD115^lo DC, produced IFN-β and TNF-α in response to GBS. IFN-β production was dependent on endosomal TLRs, most likely TLR7, whereas TNF-α production was independent of TLR7, indicating a more important role of TLR13 in this context (17).

We asked whether the heterogeneity of GM culture cells could be resolved in more than two populations based on surface marker expression. Indeed, we managed to distinguish four different populations when including CD115 and CD103 expression (population A–D, Fig. 3A). The four populations showed differences in their morphology, with an MΦ-like cell appearance in the CD115^hi CD103^lo (A) and CD115^int CD103^lo (C) populations and a rounded DC-like appearance in the CD115^hi CD103^hi (B) and CD115^lo CD103^hi (D) populations (Fig. 3A). In addition, these populations differed in qualitative and quantitative expression of F4/80, CD11c, CD135, CD86, and CD11b.

Next, we performed RNA sequencing of the defined populations during steady-state. Principal component analysis showed substantial distances between populations A and B on the one side and D on the other (Fig. 3B). In contrast, population C exhibited more transcriptional variability, indicating a transitional state. The comparison of A and D revealed 2978 DEG (Fig. 3C). Heat maps of reported key transcription factors and markers in MΦ (Csfr1, Maf, Cx3cr1) as well as DC development (Zbtb46, Flt3, Stat3, Ccr7) were highly compatible, with population A being MΦ and D being DC (Fig. 3D). When applying the same analysis in populations B and C, B showed an MΦ-like transcriptomic profile, whereas C represented an intermediate state, which was also confirmed by quantitative PCR (qPCR) (Fig. 3D). Further analysis of all DEG highlighted a rather MΦ- than DC-like transcriptome of population C, with around 530 or 2364 DEG, compared with population A or D, respectively (Supplemental Fig. 2A). To elucidate, why the MΦ-like subsets A, B, and C responded differently to GBS, we performed gene ontology analysis using the PANTHER classification system (27) and compared their DEG to that of population D. In line with DC hallmark characteristics, such as migration to lymph nodes, presentation of Ags, and regulation of T cell activation and proliferation, equivalent gene ontology (GO) terms (GO:0050900, GO:0050670, GO:0050863) were overrepresented in population D relative to populations A and B (Fig. 3E, Supplemental Fig. 2B). Interestingly, some of these terms were also overrepresented in population C (Supplemental Fig. 2B). In contrast, A, B, and C showed overrepresentation of TLR signaling genes (GO:0002224, GO:0034154) as compared with D. Furthermore, genes responsible for type I IFN production (GO:0032481) (Fig. 3E, Supplemental Fig. 2B) and subclass “positive regulation of IFN-β production” (GO:0032728, Fig. 3F) were overrepresented in A and C.

In summary, we found that GM culture gives rise to at least four cell populations with differences in morphology, surface marker expression, and cytokine response to bacteria. Whereas all CD115^hi cell populations harbored an MΦ-like transcriptomic profile, CD115^lo CD103^hi cells (population D) showed a transcriptional profile compatible with DC.

We next wondered whether the populations kept immunophenotypic flexibility despite prolonged differentiation. To test this, GM culture MΦ were sorted and stimulated with GBS over several time periods, and surface expression of CD115 and MHC II was analyzed. Under these conditions, MΦ downregulated CD115 and upregulated MCH II as well as CD86. Thus, GBS stimulation resulted in MΦ acquiring some DC properties (Fig. 4A). However, important differences between MΦ and DC remained; for example, RelB and Stat3 were transcriptionally activated in DC but not MΦ stimulated with GBS (Fig. 4B). Moreover, the typical DC surface markers CCR7 and CD135 were not substantially expressed in GBS-stimulated MΦ (Supplemental Fig. 3A). Yet, within 4 h, GBS induced a reduction in cell size and the formation of cell clusters, typical features of GM-CSF–derived DC (Fig. 4C–E). To complement analysis of DC-like features in GM-CSF culture MΦ, we examined the Ag-presentation ability, a classical property of DC. For this purpose, we sorted subsets A and D of GM-CSF–derived MΦ and M-CSF–derived MΦ. We employed CD4 T cells from spleens of OT-II mice that primarily recognize OVA peptide residues 323–339 (OVA peptide) when presented by the MHC class II molecule. The proportion of CD4 T cells undergoing cell division did not differ for cocultures with either subset A or D. However, T cell stimulation by A and D was substantially higher as compared with CD4 T cells cocultured with M-CSF–derived MΦ (Fig. 4F). Moreover, the magnitude of CD4 T cell division was not further propagated by coincubation with OVA peptide, indicating that GM culture MΦ constitutively exhibit potent Ag-presentation properties. Accordingly, CD4 T cells cultivated with populations A and D showed similarly higher Ki67 accessibility as compared with CD4 T cells in the presence of M-CSF–derived MΦ (Supplemental Fig. 3B).

These data show that GM-CSF–derived MΦ have Ag-presentation capabilities and acquire further DC-like characteristics when exposed to bacteria, indicating substantial plasticity. Yet they maintain transcriptional differences to DC.

Because the GO terms “TLR7 signaling pathway” and “positive regulation of type I IFN production,” as well as “positive regulation of IFN-β,” were overrepresented in MΦ subsets (A–C) as compared with DC (D) (Fig. 3E, 3F), we further analyzed these populations regarding their IFN response to streptococci.

First, only the MΦ subsets A and C produced IFN-β (Fig. 5A), whereas IL-10 expression was similarly observed in all four cell types (Fig. 5B). To further explore this issue, we made use of a transgenic reporter mouse, where one IFN-β allele is replaced by a luciferase encoding gene (Ifn-β^wt/Δβ-luc) (28). In full accordance with the ELISA data, activation of the reporter was induced by GBS only in MΦ and not in DC (Supplemental Fig. 4A). Interestingly, a measurable induction of IFN-α was only found in population B, which represents an intermediate MΦ phenotype (e.g., CD11c^hi) (Fig. 5C). Next, we explored whether differentiated tissue MΦ (i.e., peritoneal and dermal MΦ) exhibit an IFN-β response to GBS ex vivo and in vivo. We found peritoneal MΦ, which were cultured without GM-SCF or M-CSF and stimulated with GBS, to express mRNA of IFN-β and of its inducible gene Cxcl10 (Fig. 5D). Next, peritoneal MΦ isolated from mice, which had been infected with 10⁶ CFU GBS for 18 h, but not those from control mice, exhibited substantial CXCL10 expression (Fig. 5E, Supplemental Fig. 4B). In addition, we employed an intradermal infection model of the ear pinna of Ifn-β^wt/luc mice (16). After 24 h of infection, next to neutrophils, MΦ (CD64^hi Ly6G^neg) were recruited to the dermis, as published before (29) (Supplemental Fig. 4C). As shown in vitro, tissue MΦ sorted from the infected dermis showed increased IFN-β activity as determined by bioluminescence (Fig. 5F). Accordingly, tissue MΦ respond to GBS with IFN-β and CXCL10 both in and ex vivo, indicating an important role of IFN-β signaling in the antistreptococcal repertoire in MΦ. Furthermore, we employed GM-CSF cultures from mice with deficiency in the type I IFNR (Ifnar^−/−). Upon the stimulation with either fixed or live GBS, the production of TNF-α and IL-6 was significantly reduced compared with wt MΦ (Fig. 5G). Whereas the gene expression of Ifn-β and inos in Ifnar^−/− cells was reduced, the expression of the Ifn-β dependent gene Cxcl10 was completely lacking (Fig. 5H). In summary, these results show that IFN type I signaling in MΦ is necessary for mounting a full cytokine response to GBS.

The identity of mononuclear phagocytes is determined by terminal differentiation and plasticity. It is common understanding that their development from restricted progenitors is in principle unidirectional. It requires the passage of defined maturation checkpoints where, because of endogenous (e.g., origin-dependent) and environmental factors, lineage defining fate decisions occur (30). At the level of the most differentiated progeny (i.e., tissue MΦ and DC), plasticity allows for adaptation to the target tissue. Yet despite a high degree of adaptability, lineage defining properties are preserved. In the case of MΦ, the term polarization has been coined to capture adoption of a transcriptional program, which meets spatiotemporal dependent requirements without threatening cell identity. The concept of plasticity has recently been refined by the observation that differentiated peritoneal MΦ can adopt a relatively distant alveolar MΦ phenotype after transfer from their usual site of residence to the lung (31). Moreover, plasticity at the terminal site appears to outweigh cues inherited from the progenitor, because descendance from yolk sac/fetal liver progenitors or definitive hematopoiesis is not dominantly reflected in the tissue phenotype (32). This model is fully compatible with the concept that infiltrating monocytes can develop into MΦ or DC depending on the inflammatory status of the target tissue (4). However, the limits of plasticity are a subject of controversy. In contrast, MΦ and DC exhibit an overlapping immunophenotype. As an example, MΦ in the mouse colon show strong expression “typical” DC markers CD11c and MHC II (1). In addition, MΦ and DC share many functional properties such as Ag presentation and cytokine formation. However, these features are highly context dependent. To improve exploring the limits of plasticity within lineages, it seems essential to resolve the dynamics in identity defining properties during homeostatic and inflammatory conditions.

The findings presented here are instructive for several reasons. First, we find MΦ to be a potent source of type I IFNs when in direct contact with streptococci. A type I IFN biased program is important for immune homeostasis at mucocutaneous sites. Type I IFNs have the potential to downmodulate a host of inflammatory events (e.g., the transcription and maturation of IL-1β and the formation of IL-8, IL-12 and IL-17) (33). In our hands, type I IFN signaling is essential for a potent TNF, IL-6, and iNOS response to GBS. Accordingly, by the formation of type I IFN and by the acquisition of DC-like phenotypic properties upon contact with bacterial commensals, tissue MΦ may interlink resolution of inflammation and tissue repair with propagation of adaptive immunity.

At the beginning of extrauterine life, coexistence between colonizing, yet potentially detrimental, bacteria and human host immunity has to be established at mucocutaneous barriers. An intriguing example is the relationship of subepithelial tissue MΦ and GBS, a common intestinal colonizer and frequent cause of neonatal sepsis and meningitis (34, 35). Both in the dermis and the lamina propria of the intestine, MΦ constitute the most frequent immune cell type under homeostatic conditions. Type I IFNs are exceedingly important in keeping microorganisms in check at epithelial barriers. This is probably best understood in the context of bacterial–viral coinfections. In the mouse lung, failure to induce type I IFNs and viral control promotes bacterial bloom (36). Moreover, developmental changes in type I IFN formation, such as deficiency in CpG- and rhinovirus-induced formation of IFN-α in neonatal plasmacytoid DC and an impaired IFN-β response of monocytes from humans of >65 years, have been linked to age-related difficulties in handling viruses and local microflora (37, 38). In contrast, type I IFNs have the potential to interfere with pulmonary host defense against pneumococci (39).

Many cell types including lymphocytes, MΦ, fibroblasts, endothelial cells, and epithelial cells either form or recognize GM-CSF, making it a pleiotropic cytokine. In the context of mucocutaneous surfaces, keratinocytes are an important GM-CSF source (40). It is well established that basal levels of GM-CSF in tissue and body fluids are low at steady-state and are rapidly mounted during tissue damage inflicted by mechanical or infectious trauma (40). Accordingly, GM-CSF is a typical constituent of the microenvironment, when monocytes enter the tissue and develop into resident MΦ and DCs. In the in vitro system partially modeling this situation (i.e., when bone marrow cells with MΦ and DC progenitor potential are cultured with GM-CSF), substantial heterogeneity develops, with at least four discriminable populations, coexisting. We find CD115, CD103, CD86, and CD135, but not CD11c, F4/80, and CD206, to discriminate between MΦ and DC in GM-CSF–containing culture. Thus, CD11c expression, which was used as the sole lineage defining marker in the in vitro phagocyte response to GBS, is unsuitable for discriminating DC and MΦ (13).

It was previously shown that the endosomal TLRs 7 and 13 in mice and 7 and 8 in humans are essential for the cytokine response to RNA from streptococci and staphylococci (13, 17, 18, 41–43). It appeared that cDC mounted a TLR7-dependent IFN-β response to streptococci, as compared with MΦ, which predominantly responded with a type I cytokine response (13). In this study we find that MΦ and DC cannot be discriminated based on their mediator response to bacteria alone. Instead, MΦ are a major source of bacteria induced type I IFNs and cytokines, and the response is regulated in the single-cell level by engagement of distinct TLRs. Because the TLR7-dependent type I IFN response to GBS in MΦ is particle driven, it can be assumed to be of particular importance for the onset of bacterial tissue invasion. In contrast, MΦ mount a potent type I IFN response via cytosolic sensors via a pathway requiring proliferating bacteria. It seems remarkable that two distinct TLRs, namely TLR 7 and 13, recognize structurally very related bacterial ligands, share phagosomal localization, and similarly engage signaling intermediates such as MyD88 and TAK1 (44) yet induce very distinct cytokine and IFN programs in MΦ. It is tempting to speculate that cooperation of these TLR systems has developed to allow for fine tuning of MΦ autonomous antibacterial immunity by steering individual receptor expression levels.

Notably, the landmark paper by Helft et al. (5) did not find LPS to substantially alter the immunophenotype of GM-CSF–derived MΦ. Our data markedly extend this model, because uptake of whole bacteria leads to a convergence of MΦ with DCs in biologically relevant immunophenotypic properties, like expression of the M-CSFR, cell shape, and clustering. Thus, it appears that GM-CSF induces extreme plasticity in tissue MΦ. MΦ have the potential to change their immunophenotype and morphology in close approximation to that of cDCs when in contact with mucocutaneous pathobionts, although differences between MΦ and DC persist. Accordingly, MΦ may morphologically and functionally tighten their interlinkage with the adaptive immune response after direct contact with bacteria. This induced plasticity likely plays a mechanistic role in mucocutaneous surveillance, which is associated with sporadic contact to invading bacteria.

The authors are indebted to Anita Imm and Jan Bodinek-Wersing from the Lighthouse core facility of the University Medical Center Freiburg for outstanding technical assistance.

The authors have no financial conflicts of interest.