Independent of Plasmacytoid Dendritic Cell (pDC) infection, pDC Triggered by Virus-Infected Cells Mount Enhanced Type I IFN Responses of Different Composition as Opposed to pDC Stimulated with Free Virus Free

Type I IFN responses are induced within hours after infection with many different pathogens and may confer protection against lethal disease (1). The type I IFN family consists of 14 IFN-α subtypes, IFN-β, IFN-ε, IFN-ω, IFN-κ, and IFN-τ (1). In cell culture, basically all infected cell types are capable to mount IFN responses, whereas in vivo plasmacytoid dendritic cells (pDC) constitute a major IFN-producing cell subset (2, 3). IFN responses are induced by triggering of pattern recognition receptors such as TLRs, RIG-I–like helicases (RLH), C-type lectin-like receptors, as well as several intracellular DNA sensors and other not yet identified receptors (4).

In pDC, the mechanism of IFN induction is different compared with many other immune cell subsets. This is because pDC constitutively express IFN regulatory factor (IRF)-7 and therefore are qualified to mount swift IFN responses (5). Furthermore, upon triggering with several different stimuli, such responses may be independent of IFN receptor feedback (5, 6). Unlike other DC subsets or monocytes, human pDC express endosomal TLR7 and TLR9 and are therefore preferentially stimulated by viral and bacterial nucleic acids (7). Whereas TLRs as the major signaling platform for triggering IFN-I responses in pDC are well established, the relevance of RLHs, which may be induced upon RNA stimulation, is less clear (8). Despite their uncertain role in pDC, RLHs are the key signaling platforms for IFN induction of myeloid cells such as myeloid DC (mDC) and macrophages (9, 10). Thus, pDC use a different setup of pattern recognition receptors to mount distinct innate immune responses compared with myeloid cells.

Although in vivo pDC represent a rare cell subset compared with other immune cells, they are capable of conferring systemic IFN responses upon various different viral infections (2, 11). Strong IFN responses can even be observed upon treatment with replication-deficient or inactivated viruses (12). These facts suggest the presence of an amplification mechanism that ensures IFN triggering of pDC. Although pDC are not productively infected by obligatory cell type–specific viruses such as hepatitis C virus (HCV), they mount IFN responses upon triggering by direct cell-to-cell contact with HCV-infected hepatocytes (13). Under such conditions, hepatocyte-derived virus-free exosomes containing viral RNA shuttling to pDC trigger IFN expression in pDC (14, 15). Interestingly, following infection of mice with the cytopathic vesicular stomatitis virus (VSV), a decrease of pDC numbers was detected that was dependent on IFN expression of pDC (16). However, it is unclear whether direct infection with a cytopathic virus is needed for pDC to mount effective type I IFN responses.

In the present study, we show that pDC are remarkably resistant to productive infection with RNA-encoded VSV, but still mount substantial IFN responses upon VSV treatment. In in vitro as well as in vivo studies such IFN responses are primarily mediated by uninfected pDC. Furthermore, stimulation of bone marrow–derived (BM-)pDC with VSV-infected myeloid cells triggered increased IFN levels compared with free virus. Stimulation experiments with newly bred double IFN reporter mice that express GFP under the promoter of IFN-α6 and YFP together with IFN-β revealed the induction of differentially skewed IFN responses, depending on whether pDC were stimulated by VSV-infected cells or by free virus. These experiments implied a general mechanism of IFN induction in pDC relying on cellular crosstalk with virus-infected myeloid cells.

IFN-β reporter mice (messenger of IFN-β [MOB]) were described before (17). To obtain IFN-α6 and IFN-β double reporter mice (messenger of IFN-β and IFN-α [MOBA]), MOB mice were intercrossed with IFN-α6–GFP reporter mice (messenger of IFN-α [MOA]) (10). CARDIF^−/− mice (18) and MyD88^−/− mice were described before (19). To obtain MyD88^−/−TRIF^−/− mice, MyD88^−/− mice were intercrossed with TRIF^−/− mice (20). Double-knockout MyD88^−/−TRIF^−/− mice were further intercrossed with CARDIF^−/− mice to obtain triple-knockout MyD88^−/−TRIF^−/−CARDIF^−/− mice. Type I IFN receptor-deficient mice (IFNAR^−/−) (21) have been 20-fold backcrossed on the C57BL/6 background (22). IFN-β–deficient mice (IFN-β^−/−) (23), IRF-1–deficient mice (IRF-1^−/−) (24), and TLR7-deficient mice (TLR7^−/−) (25) were described before. All mice were bred under specific pathogen-free conditions at the central animal facility of the TWINCORE or the Helmholtz Centre for Infection Research, Braunschweig, Germany. C57BL/6 mice were purchased from Harlan Winkelmann. Mouse experimental work was carried out using 8- to 20-wk-old mice. All animals were handled in strict compliance with regulations of the German Animal Welfare Law. The protocol was approved by the Niedersächsisches Landesamt für Verbraucherschutz und Lebensmittelsicherheit (identification no. 09/1655).

VSV serotype Indiana (Mudd-Summers isolate) was originally obtained from D. Kolakofsky (University of Geneva, Geneva, Switzerland). The VSV variant with mutated matrix protein 2 (VSV-M2) is a natural variant of the VSV wild-type strain HR (26, 27). The eGFP-expressing variants enhanced GFP–expressing VSV (VSVeGFP) (28), expressing eGFP upon VSV replication in infected cells, and VSV-G–GFP (29), expressing GFP as a fusion protein with VSV-G, were provided by J.K. Rose. VSV-expressing red fluorescent protein (VSV-RFP) was provided by S.P.J. Whelan (30). All viruses were routinely propagated on BHK-21 cells and titrated on Vero cells. Virus stocks were prepared as culture supernatants 72 h postinfection with low multiplicity of infection (MOI; 0.01).

For in vitro generation of BM-DC, bone marrow cells were isolated by flushing femur and tibia of mice with RPMI 1640 medium supplemented with 10% FCS, 10 mM HEPES, 1 mM sodium pyruvate, 2 mM GlutaMAX (Life Technologies), 100 U/ml penicillin (Life Technologies), 100 μg/ml streptomycin (Life Technologies), and 0.1 mM 2-ME. RBCs were lyzed from bone marrow cell preparations using RBC lysing buffer (Sigma-Aldrich). Bone marrow cells were cultivated for 8 d at a density of 1 × 10⁶/ml (mDC) or 2 × 10⁶/ml (pDC) supplemented with 100 ng/ml GM-CSF (mDC) or Flt-3L (pDC) (both PeproTech). The medium was changed once during BM-pDC culture and three to four times for BM-mDC cultures by replacing two-thirds of the medium with fresh cytokine-supplemented medium.

For stimulation, in vitro–differentiated BM-pDC and BM-mDC were seeded at 1 × 10⁶ cells/ml in 24-well culture plates in RPMI 1640 with 10% FCS. Cells were treated at an MOI of 1 or 10 with the indicated viruses or stimulated with 100 nmol CpG-containing oligodeoxynucleotide 2216 (InvivoGen). For cocultivation experiments of BM-pDC with infected cells, BM-mDC were treated with VSV (MOI of 10), incubated for 18 h, and then extensively washed. In a next step, 1 × 10⁶ BM-pDC were mixed with 1 × 10⁶ infected BM-mDC treated as indicated. After 18 h of stimulation, culture supernatants were collected and analyzed by ELISA, or cells were harvested for flow cytometric analysis. Because minor variations may occur in experiments with differentiated primary bone marrow cells, all controls were included for every single experiment. For RNase and DNase digestion, either 1 × 10⁶ VSV PFU or VSV-infected BM-mDC were incubated with 0.5 μg RNase (Roche) for 15 min, or 1 U DNase (Roche) in the recommended buffers for 30 min at 37°C. Subsequently, stimuli were added as described.

Magnetic enrichment of CD11c⁺ splenocytes was carried out using CD11c microbeads and an autoMACS Pro separator following the manufacturer’s protocol (Miltenyi Biotec). BM-pDC and BM-mDC were stained with anti–Siglec-H, anti-CD11b, and anti-CD11c Abs. All lineage Abs were purchased from eBioscience or BD Pharmingen. Analysis of cell viability is shown in Supplemental Fig. 1. FACS sorting of pDC was carried out after Siglec-H staining of Flt-3L culture cells and subsequent sorting with a FACSAria (BD Biosciences) or MoFlo-XDP (Beckman Coulter) device. Anti-calreticulin Ab was purchased from StressMarq Biosciences. Cells were analyzed by flow cytometry (LSR II SORP; BD Biosciences) and data were analyzed using the FlowJo software. For analysis of GFP⁺/YFP⁺ cells, LSR II SORP with YFP 542/27 and GFP 510/20 filters (AHF Analysentechnik) was used. IFN-α was detected in cell culture supernatants using an ELISA method (eBioscience).

To study the cellular mechanism of virus-induced IFN responses, we isolated bone marrow from wild-type mice and in vitro–differentiated BM-pDC and BM-mDC and stimulated the cells with VSV at an MOI of 1 or 10. After 18 h incubation, cell-free supernatant was harvested and the IFN-α content was determined by an ELISA method. Upon stimulation with VSV at an MOI of 1, BM-pDC and BM-mDC mounted similar IFN-α responses (Fig. 1A). As reported previously (27), at an MOI of 10 primarily BM-pDC, but not BM-mDC, were triggered to mount IFN responses (Fig. 1A). Next, we tested VSVeGFP (28), which at an MOI of 1 induced even stronger IFN-α responses in mDC compared with pDC (Fig. 1A). At an MOI of 10, VSVeGFP stimulation of BM-mDC did not induce IFN-α as observed before in experiments with wild-type VSV (Fig. 1A). Of note, VSVeGFP always induced enhanced IFN responses compared with wild-type VSV. This can be explained by the attenuation of VSVeGFP caused by the eGFP transgene expression.

To address whether triggering of IFN responses was associated with productive infection of the IFN-producing cells, BM-mDC and BM-pDC were infected with VSVeGFP at an MOI of 1 or 10, and after 6, 12, 18, and 24 h incubation eGFP expression was monitored cytofluorometrically. At an MOI of 1, pDC did not show significant eGFP expression at the time points tested (Fig. 1B). Even at an MOI of 10 only moderately enhanced eGFP expression was detected that was slightly augmented after 18 h incubation (Fig. 1B). On the contrary, already 6 h after VSVeGFP infection at an MOI of 1, 33% of BM-mDC were GFP⁺ and after 12 and 18 h incubation this value increased up to 77% (Fig. 1C). At an MOI of 10, after 12 and 18 h of VSVeGFP infection >90% of BM-mDC were eGFP⁺ (Fig. 1C). Twenty-four hours postinfection the percentage of eGFP⁺ pDC and mDC again decreased because of cytopathic effects (Fig. 1B, 1C). To verify that eGFP expression of cells was conferred by productive VSVeGFP infection, we UV-irradiated VSVeGFP with increasing dosages prior to treatment of BM-mDC. Upon BM-mDC stimulation with VSVeGFP irradiated with a UV dose of 750 mJ/cm², no eGFP signals were detected anymore in BM-mDC (Supplemental Fig. 2). These observations indicated that under the conditions tested pDC were largely resistant to productive VSV infection, whereas mDC were readily infected.

Of note, resistance of pDC to VSV infection was only partially conferred by type I IFN receptor stimulation, as indicated by type I IFN receptor–deficient (IFNAR^−/−) BM-pDC showing only moderately enhanced infection compared with wild-type pDC (Fig. 1D). In contrast, at an MOI of 1 and 10 IFNAR^−/− BM-mDC were infected virtually quantitatively (Fig. 1D), whereas wild-type BM-mDC were only partially infected (Fig. 1C). These results demonstrated that upon IFNAR triggering BM-mDC showed enhanced resistance, whereas in pDC in addition to IFNAR some other mechanism played a role. Interestingly, IRF-1–deficient BM-pDC showed ∼4-fold enhanced susceptibility to VSVeGFP infection, as indicated by 62% eGFP⁺ IRF-1^−/− pDC and 16% wild-type BM-pDC upon VSVeGFP treatment (Fig. 1E). Thus, in pDC an IRF-1–dependent mechanism mainly conferred resistance, instead of IFNAR triggering.

To next address whether TLR and/or RLH signaling mediated VSV sensing, BM-pDC and BM-mDC were generated from MyD88^−/−TRIF^−/− deficient mice devoid of TLR signaling, or CARDIF^−/− mice devoid of RLH signaling. Whereas VSVeGFP-infected MyD88^−/−TRIF^−/− and IFNAR^−/− BM-pDC showed significantly enhanced eGFP expression when compared with wild-type controls, no enhanced eGFP expression was detected upon infection of CARDIF^−/− BM-pDC (Fig. 2A, upper panel). On the contrary, CARDIF^−/− BM-mDC showed similarly enhanced eGFP expression as IFNAR^−/− BM-pDC, whereas VSVeGFP-infected MyD88^−/−TRIF^−/− BM-mDC showed similar eGFP expression as did wild-type mDC (Fig. 2A, lower panel). These observations indicated that in BM-pDC TLR, but not RLH, signaling played the key role to limit VSV infection, whereas in BM-mDC RLH, but not TLR, signaling was critically involved.

To next dissect the mechanism of VSV-triggered IFN induction, BM-pDC and BM-mDC were generated from wild-type, IFNAR^−/−, MyD88^−/−TRIF^−/−, and CARDIF^−/− mice. The cells were treated with VSV or VSVeGFP at an MOI of 1 or 10 and after 18 h of incubation cell-free supernatant was tested for IFN-α. Wild-type BM-pDC stimulated with VSV at an MOI of 1 mounted IFN-α responses of ∼0.9 ng/ml, whereas at an MOI of 10 even higher IFN-α levels of 1.2 ng/ml were reached (Fig. 2B). CARDIF^−/− BM-pDC mounted similar IFN-α responses as did wild-type BM-pDC, whereas MyD88^−/−TRIF^−/− and IFNAR-deficient BM-pDC did not produce IFN (Fig. 2B). Comparable results were obtained when pDC were stimulated with VSVeGFP (Fig. 2B). Thus, pDC sense VSV and VSVeGFP in a TLR-dependent manner.

Previously it was shown that expression of a fully functional VSV matrix protein conferred induction of reduced IFN responses (27). Correspondingly, stimulation with VSV-M2 exhibiting an M15R point mutation within the matrix protein induced an increased IFN-α response in wild-type BM-pDC when compared with wild-type VSV (Fig. 2B). Similar to VSV and VSVeGFP, VSV-M2 induced IFN-α responses in CARDIF^−/− BM-pDC, whereas surprisingly and unlike the other two virus strains tested, VSV-M2 triggered IFN-α responses also in MyD88^−/−TRIF^−/− BM-pDC (Fig. 2B). To verify that IFN responses detected in VSV-M2–treated Flt-3L cultures were indeed derived of pDC, Siglec-H⁺ pDC were FACS sorted and further analyzed (Fig. 2C). VSV-M2 triggered reduced but significant IFN-α responses in sorted MyD88^−/−Trif^−/− deficient pDC, indicating that VSV-M2 was not exclusively sensed via TLR in pDC. To analyze whether in the case of VSV-M2 RLH and TLR signaling redundantly interacted to induce IFN responses, MyD88^−/−TRIF^−/−CARDIF^−/− BM-pDC lacking RLH as well as TLR signaling were tested. Indeed, RLH/TLR-deficient BM-pDC did not mount IFN responses upon treatment with VSV, VSVeGFP, or VSV-M2 (Fig. 2B). Similarly, IFNAR-deficient BM-pDC were not induced by any of the three virus strains to mount IFN responses (Fig. 2B). Thus, whereas TLR signaling was needed in pDC to confer VSV- or VSVeGFP-stimulated IFN responses, in the case of VSV-M2, TLR and RLH signaling redundantly interacted to induce IFN. Furthermore, IFNAR feedback is essential not to protect pDC, but to confer positive feedback to promote IFN induction. In line with the above data, TLR7^−/− BM-pDC were not able to produce IFN-α in response to VSV or VSVeGFP stimulation, suggesting that VSV-derived RNA was sensed by endosomal TLR7 (Fig. 2D). Of note, and in agreement with the above results, upon stimulation with VSV-M2, TLR7^−/− BM-pDC mounted similar IFN responses as did wild-type pDC (Fig. 2D).

In a next step, responses of myeloid cells were studied. Wild-type and MyD88^−/−TRIF^−/− BM-mDC stimulated with either VSV or VSVeGFP mounted comparable IFN-α responses (Fig. 2E). In contrast to pDC, CARDIF-deficient BM-mDC did not mount IFN-α responses upon stimulation with either of the three virus strains (Fig. 2E). Furthermore, MyD88^−/−TRIF^−/−CARDIF^−/− and IFNAR^−/− mDC did not produce detectable IFN-α (Fig. 2E). Thus, in BM-mDC VSV, VSVeGFP, and VSV-M2 triggered IFN responses in an RLH-dependent manner. Similar to BM-mDC, BM-macrophages treated with VSV or VSVeGFP also mounted RLH-dependent and TLR-independent IFN responses (data not shown). In conclusion, these results are compatible with the model that pDC sense VSV primarily TLR7-dependently and RLH-independently, whereas mDC and macrophages are triggered upon virus infection RLH-dependently and TLR7-independently to mount IFN responses. However, because in pDC VSV-M2 is sensed by TLR as well as RLH, VSV-M–mediated shutdown of nuclear RNA export seems to particularly efficiently inhibit RLH-dependent IFN triggering of pDC.

To address whether virus infection was a prerequisite of IFN induction, pDC and mDC were generated from bone marrow of MOB mice expressing YFP in conjunction with IFN-β. Upon VSVeGFP treatment and subsequent incubation for 18 h, within BM-pDC and BM-mDC cultures, Siglec-H⁺ pDC and CD11c⁺CD11b⁺ mDC, respectively, were analyzed for YFP and eGFP expression. At an MOI of 10, >90% of YFP⁺ pDC were GFP⁻, indicating that most IFN-β–producing pDC were not infected (Fig. 3A, upper panels). On the contrary, most YFP⁺ mDC were also GFP⁺, indicating that infected mDC were the main IFN-β producers (Fig. 3A). To address whether pDC might have been infected without subsequent initiation of viral gene expression and viral replication, a second VSV reporter virus was used expressing a fusion protein consisting of VSV-G and GFP on the particle surface (VSV-G–GFP). Such virus may give rise to GFP fluorescence independent of productive infection and thus may decorate infected cells without having passed a full replication cycle (29). Upon VSV-G–GFP stimulation of BM-pDC and BM-mDC, similar results were obtained as observed before with VSVeGFP, that is, VSV-G–GFP infection resulted in slightly reduced, but overall comparable, percentages of YFP⁺GFP⁻ pDC as well as YFP⁺GFP⁺ mDC (Fig. 3B). To verify the above findings in vivo, MOB mice were infected with 2 × 10⁶ PFU VSVeGFP i.v. and 24 h later YFP- and eGFP-expressing cells were analyzed from spleen. Whereas a proportion of 1.6% YFP⁺ IFN-β–producing cells was characterized to be Siglec-H⁺CD11c^lo pDC, virtually no pDC were found within the GFP⁺ population (Fig. 3C). Hence, direct VSV infection is a prerequisite for myeloid cells, but not for pDC, to mount IFN responses. It is noteworthy that in all infection experiments, comprising in vitro studies as well as in vivo analyses, a surprisingly small proportion of cells produced IFN.

Because IFN-producing pDC usually are not infected, we next sought whether free virus and infected cells similarly induced IFN responses. To this end, BM-pDC were stimulated either with free virus or with BM-mDC, which were infected with VSV at an MOI of 10, incubated for 18 h, and then intensively washed. Interestingly, cocultures of BM-pDC with infected mDC induced significantly enhanced IFN responses when compared with free virus (Fig. 4A). Similarly, BM-pDC stimulated with supernatants of VSV-infected BM-mDC showed significantly enhanced induction of IFN-α expression (Supplemental Fig. 3). Of note, in experiments in which free VSV or VSVeGFP was added at an MOI of 30, no further enhancement of IFN responses was obtained (Fig. 4B). Thus, the absolute quantity of free virus was not limiting the magnitude of IFN responses mounted by pDC. To address whether enhanced virus titers derived of infected mDC cultures triggered increased IFN responses, virus-infected BM-mDC were coincubated with BM-pDC in the presence of 1% VSV neutralizing hyperimmune serum, which eliminated free virus. Neutralization of free virus in the supernatant did not reduce the enhanced induction of IFN responses by infected BM-mDC, although absolute amounts of induced IFN responses were reduced compared with controls devoid of hyperimmune serum (Fig. 4C). Furthermore, the magnitude of IFN responses was similar when either infected IFN-β–deficient or IFN-β–competent BM-mDC were used to stimulate BM-pDC (Fig 4D). Comparable results were obtained when wild-type BM-pDC were stimulated with VSV-infected IFNAR^−/− BM-mDC, which also induced elevated amounts of IFN (data not shown). Thus, mDC-derived IFN did not play a role in enhancing IFN responses. Recently it was shown that infected hepatocytes stimulated pDC by hepatocyte-derived exosomes containing viral RNA (15). Therefore, we aimed to further characterize the properties of the IFN-stimulating factor produced by infected cells. To this end, BM-mDC infected with VSVeGFP were incubated with DNase or RNase prior to coincubation with BM-pDC. Under such conditions neither DNase nor RNase treatment impaired the enhanced IFN-inducing properties of infected mDC (Fig. 4E). Interestingly, it was shown in VSV-infected drosophila that autophagy was essential to mediate anti-VSV immunity (31). To analyze infection-induced autophagy, BM-pDC were stained for surface-exposed calreticulin 18 h after treatment with VSVeGFP. These data revealed increased surface-exposed calreticulin in eGFP⁺ cells compared with eGFP⁻ uninfected BM-pDC, indicating that autophagy was selectively induced in infected pDC (Fig. 4F). To assess whether autophagy of BM-pDC was involved in the induction of IFN responses, pDC were treated with the PI3K inhibitor wortmannin during incubation with VSV. Of note, wortmannin treatment inhibited the induction of IFN-α during stimulation either with free virus or VSV-infected mDC (Fig. 4G). Similar data were obtained when the autophagy inhibitor 3MA was used (data not shown). To study whether autophagy was also involved in the production of the pDC stimulating factor released by infected mDC, BM-mDC were treated with wortmannin during VSV infection and prior to cocultivation with pDC. To our surprise, such wortmannin-treated BM-mDC still normally triggered BM-pDC to mount enhanced IFN responses (Fig. 4F). Thus, autophagy of pDC was essential for VSV-mediated triggering of IFN responses in pDC, whereas autophagy of infected cells was not required to release pDC-stimulating factors.

To study IFN responses of pDC induced by free virus or by VSV-infected BM-mDC on the single cell level, IFN-α6 and IFN-β double reporter mice (MOBA) were bred. In cells derived of such mice, YFP⁺ IFN-β–expressing cells, GFP⁺ IFN-α6–expressing cells, and YFP⁺GFP⁺ IFN-β and IFN-α6–expressing cells can be discriminated. Indeed, upon CpG stimulation of MOBA, Siglec-H⁺ BM-pDC, subsets of IFN-β– or IFN-α6–expressing pDC, as well as double-positive cells were detected (Fig. 5A). Stimulation of MOBA BM-pDC with free VSV-M2 also induced expression of either IFN-β, IFN-α6, or both cytokines on the single-cell level (Fig. 5A). Of note, IFN-α6 was also not induced in directly infected BM-pDC but rather in uninfected cells, as tested by infection of MOA pDC with VSV-RFP (Supplemental Fig. 4). Because the percentages of subsets of IFN-expressing pDC differed depending on the stimulus used (Fig. 5A), we next compared IFN responses being induced upon treatment with free VSV and VSV-infected mDC. Interestingly, upon stimulation of MOBA BM-pDC with VSV-infected BM-mDC, the percentage of IFN-β⁺ and IFN-β⁺/IFN-α6⁺ double-positive pDC subsets was increased when compared with stimulation experiments using free virus (Fig. 5B). Thus, stimulation of pDC with infected mDC instead of free virus influenced the quantity as well as the quality of type I IFN responses.

Type I IFN responses constitute one very potent antiviral defense mechanism that can be induced either locally or systemically. Upon in vivo infection, systemic IFN responses are presumably contributed by pDC (2, 11). IFN triggering may be conferred by TLR- and/or RLH-dependent signaling platforms and possibly by other not yet identified mechanisms (4). In the present study we found that upon VSV treatment BM-pDC produced massive IFN responses in a TLR7-dependent manner. Compared with pDC, mDC showed reduced IFN responses, whereas IFN induction was dependent on RLH signaling. Thus, depending on which cell subset was triggered by VSV, different signaling platforms played the key role. Strikingly, in pDC, VSV-M2 was not exclusively sensed by TLRs. Instead, CARDIF^−/− pDC showed similar IFN responses as did wild-type pDC, whereas MyD88^−/−TRIF^−/− pDC showed even enhanced IFN responses. Importantly, depletion of MyD88/Trif/Cardif abolished VSV-M2–induced IFN responses completely. RLH-dependent sensing of RNA-encoded viruses has previously been discussed as a unique hallmark of myeloid cells (9, 10). However, it was shown before that RLH might be upregulated in response to stimulation with pattern recognition receptor ligands or pathogens (32, 33). In this context, recently Stone et al. (34) described the induction of type I and type III IFNs by an HCV-encoded RNA in a pDC-like cell line (GEN2.2-pDC) to be RLH-dependent. In the present study, we verified that VSV-M2 induced significantly enhanced IFN responses compared with wild-type VSV. The underlying mechanism was discussed to be contributed by increased export of cellular mRNA including IFN transcripts (27). Our data imply that stimulatory RNA might be induced in infected pDC, which induces cytosolic RNA-specific pattern recognition receptors such as RLH and thus triggers pDC TLR independently.

Our studies revealed that although in BM-pDC cultures only a minority of the cells are infected, high amounts of IFN are produced. By using reporter mouse lines we found in in vitro and in vivo studies that on the single cell level 1) very few pDC contribute to the production of high amounts of IFN, and that 2) IFN-producing pDC are not infected. In stark contrast, primarily VSV-infected myeloid cells produced IFN, and the overall amount of IFN was dramatically reduced compared with pDC. VSV-encoded matrix protein leads to an effective nuclear shutdown in infected cells (35). In contrast, VSV-M2 showing reduced matrix protein–mediated retention of nuclear mRNA triggered increased IFN production of pDC in a TLR-independent manner. It is of major interest whether upon VSV-M2 stimulation also very few infected pDC would mount IFN responses as observed with wild-type VSV.

The important role of pDC in antiviral defense makes them a critical target for evasion strategies of different pathogens. Besides the modulation of signaling pathways of pDC, some viruses such as arenavirus, lymphocytic choriomeningitis virus, and Lassa virus even productively infect pDC (36). Nevertheless, we found that pDC were remarkably well protected against VSV infection. Interestingly, this protection of pDC was not exclusively conferred by IFNAR feedback signaling but additionally was mediated by IRF-1 and presumably other mechanisms. IRF-1 has earlier been described to confer protection against different infections such as reoviruses and Brucella (37, 38).

Despite the fact that direct virus stimulation of pDC leads to IFN induction, only a minority of cells within BM-pDC cultures are infected. Moreover, infected myeloid cells (or their supernatants) constitute a more profound stimulus of pDC-derived IFN when compared with stimulation by free virus. Interestingly, the enhanced stimulation of BM-pDC was not mediated by IFN derived from infected BM-mDC because neither IFN-β^−/− nor IFNAR^−/− BM-mDC showed reduced stimulation of BM-pDC. Presumably, the pDC stimulus is conferred by subcellular material released from dying cells upon infection. Of note, also with other viruses such as modified vaccinia virus Ankara, we found that infected cells triggered stronger IFN responses than did free virus (data not shown).

Takahashi et al. (13) previously described that human pDC are stimulated by HCV-infected hepatocytes to mount IFN responses, while being resistant to direct HCV infection. Just recently it was found that exosomes derived from hepatitis B virus–infected cells mediate a cell-to-cell transmission of IFN-α–inducing signals (39), and exosomes might even contain stimulatory RNA from HCV-infected cells, which is sensed TLR7 dependently in pDC (15). Also just recently Python et al. (14) showed that swine fever virus–infected SK-6 cells stimulated porcine pDC to mount stronger IFN-α responses compared with free swine fever virus. Interestingly, in the same study the viral RNase E^rns of swine fever virus efficiently inhibited sensing of infected cells by pDC. The sensing of VSV as well as of VSV-infected cells was blocked by wortmannin or 3-MA treatment of BM-pDC. These observations indicated that independent of whether pDC were stimulated by free virus or infected cells, the induction of IFN responses was dependent on PI3K activity of pDC. Among others, PI3K is closely linked to different signaling pathways such as autophagy and stress-mediated responses. Recently, it was shown that in Drosophila autophagy directly conferred antiviral effects against VSV (31). In these flies VSV-G was sensed by pattern recognition receptors and induced autophagy that efficiently inhibited viral replication. We showed by determination of surface-exposed calreticulin that VSV infection induced autophagy also in BM-pDC. In this context, our data further supported the model that stimulatory exosomes that are produced by infected cells might enter endocytotic pathways of pDC to induce protective IFN independently of direct infection.

Different type I IFN isotypes confer signaling via the common IFN receptor IFNAR. Nevertheless, different type I IFN isotypes may bind with different affinities to IFNAR as exemplified for IFN-α and IFN-β (40), implying a differential triggering of specific patterns of regulation of IFN-induced genes, and different outcomes of antiviral immunity, depending on the IFN composition induced. By using novel double reporter mice we studied IFN-α6 and IFN-β induction on the single-cell level. We found that stimulation of pDC with infected cells resulted in another distribution of IFN-α6⁺, IFN-β⁺, and IFN-α6⁺IFN-β⁺ cell subsets than observed after stimulation with free virus. Previously, IFN-β⁺ cells were phenotypically described as a specialized subpopulation of TNF and inducible NO synthase–producing DCs (41). It will be a matter of future research to determine markers that qualify subsets of pDC and mDC to respond to virus stimulation and to mount IFN responses. Although total IFN-α expression was massively increased upon stimulation with infected cells, as indicated by the IFN-α levels detected in cell-free supernatant, the GFP expression in MOBA BM-pDC was not induced equivalently. One explanation for this is that IFN-α6 is not the major IFN-α subtype being induced upon stimulation with VSV-infected cells. Indeed, Barchet et al. (2) showed that depending on whether splenocytes or mouse embryofibroblasts were analyzed, particularly IFN-α2 and IFN-α5 were induced whereas IFN-α4 was found only rarely.

In conclusion, our results imply a common mechanism of pDC stimulation by virus-infected cells that induces a distinct quality and quantity of IFN responses compared with free virus and that might be conserved across different species. The differential composition of IFN responses, depending on whether stimulated by either free virus or virus-infected cells, might facilitate orchestration of suitable IFN response against cell-free cytopathic viruses or intracellular pathogens.

We thank J.K. Rose for providing GFP-expressing VSV strains and S.P.J. Whelan for VSV-RFP. We thank S. Akira, B. Beutler, S. Weiß, A. Kröger, and D. Bruder for providing different mouse strains used in this study (IFN-α6-GFP, Trif^−/−, IFN-β^−/−, IRF-1^−/−, and TLR7^−/− mice, respectively). We thank J. Heinrich for expert help with mouse breedings and screenings, and in particular for providing MyD88^−/−TRIF^−/−CARDIF^−/− mice.

The authors have no financial conflicts of interest.