key: cord-0000198-hx77tpbt authors: Fejer, György; Drechsel, Lisa; Liese, Jan; Schleicher, Ulrike; Ruzsics, Zsolt; Imelli, Nicola; Greber, Urs F.; Keck, Simone; Hildenbrand, Bernd; Krug, Anne; Bogdan, Christian; Freudenberg, Marina A. title: Key Role of Splenic Myeloid DCs in the IFN-αβ Response to Adenoviruses In Vivo date: 2008-11-14 journal: PLoS Pathog DOI: 10.1371/journal.ppat.1000208 sha: 9c33102dc55e225f5836f9cf9c9d5f5b918820d8 doc_id: 198 cord_uid: hx77tpbt The early systemic production of interferon (IFN)-αβ is an essential component of the antiviral host defense mechanisms, but is also thought to contribute to the toxic side effects accompanying gene therapy with adenoviral vectors. Here we investigated the IFN-αβ response to human adenoviruses (Ads) in mice. By comparing the responses of normal, myeloid (m)DC- and plasmacytoid (p)DC-depleted mice and by measuring IFN-αβ mRNA expression in different organs and cells types, we show that in vivo, Ads elicit strong and rapid IFN-αβ production, almost exclusively in splenic mDCs. Using knockout mice, various strains of Ads (wild type, mutant and UV-inactivated) and MAP kinase inhibitors, we demonstrate that the Ad-induced IFN-αβ response does not require Toll-like receptors (TLR), known cytosolic sensors of RNA (RIG-I/MDA-5) and DNA (DAI) recognition and interferon regulatory factor (IRF)-3, but is dependent on viral endosomal escape, signaling via the MAP kinase SAPK/JNK and IRF-7. Furthermore, we show that Ads induce IFN-αβ and IL-6 in vivo by distinct pathways and confirm that IFN-αβ positively regulates the IL-6 response. Finally, by measuring TNF-α responses to LPS in Ad-infected wild type and IFN-αβR(−/−) mice, we show that IFN-αβ is the key mediator of Ad-induced hypersensitivity to LPS. These findings indicate that, like endosomal TLR signaling in pDCs, TLR-independent virus recognition in splenic mDCs can also produce a robust early IFN-αβ response, which is responsible for the bulk of IFN-αβ production induced by adenovirus in vivo. The signaling requirements are different from known TLR-dependent or cytosolic IFN-αβ induction mechanisms and suggest a novel cytosolic viral induction pathway. The hypersensitivity to components of the microbial flora and invading pathogens may in part explain the toxic side effects of adenoviral gene therapy and contribute to the pathogenesis of adenoviral disease. Adenoviruses (Ads) cause mild disease in humans, but are hazardous pathogens in immuno-compromised individuals [1] . Human Ads are dsDNA viruses grouped into six species. Species A, C, D, E, and F and species B Ads use different infectious entry pathways [2] . Human Ads enter mouse cells and express their early genes; however, the virus genome is not replicated and no viral progeny is made during the infection of mouse cells in vitro or in vivo [3, 4] . Furthermore, early viral gene expression can be abolished by UV-inactivation and well-defined mutants with defects of viral early genes or viral entry are available [3, 5] . Thus, the effects elicited by different components of the virus-host interaction preceding viral replication can be accurately evaluated. Ads transduce many different cell types and can be produced in vitro in sufficiently high amounts for in vivo administration. While these properties make them attractive for gene therapy applications, they can also trigger a severe systemic toxic reaction [6, 7] . Upregulation of inflammatory mediators, including cytokines and chemokines such as IL-1, IL-6, IL-8, IL-12, macrophage inhibitory protein-1/2, tumor necrosis factor-a (TNF) and recently also type I IFN has been observed in experimental and clinical infections with wt as well as with recombinant Ads [6, 7, 8, 9, 10] . Type I IFNs represent one of the host's most important antiviral defense mechanisms. The type I IFN family comprises different IFN-a subtypes, a single IFN-b and other less well characterized proteins [11] . All IFN-a species and IFN-b interact with the same IFN-ab cellular receptor, the activation of which mediates a wide range of direct and indirect innate antiviral or antimicrobial effects and modulates the antiviral adaptive immune response [12, 13, 14] . At present, two main mechanisms of type I IFN induction by viruses resulting from the extracytoplasmic or cytoplasmic virus recognition, respectively, are known [12, 13, 14, 15] . The extracytoplasmic induction is initiated by triggering the surfaceexpressed transmembrane protein toll-like receptor (TLR) 4 with certain non-nucleic viral constituents [16, 17, 18] or upon recognition of viral nucleic acids in the endosomes of specialized cells (dendritic cells and macrophages) via different members of the TLR family. These include TLR3, TLR7/TLR8 and TLR9, sensing dsRNA, ssRNA and CpG DNA, respectively. For IFN-ab induction TLR3 and TLR4 signal through the adaptor molecule TIR domain-containing adaptor protein inducing interferon b (TRIF). This results in the activation of interferon regulatory factor (IRF)-3. TLR7, 8 and 9 signal through the adaptor molecule Myeloid differentiation factor 88 (MyD88). An important result of the MyD88-mediated pathway is the activation of IRF-7 (but not of IRF-3), which together with the transcription factors NF-kB and AP-1 initiates the induction of both the IFN-a and IFN-b genes [13, 15] . This induction pathway is responsible for the strong, early IFN-ab response to several replicating and inactivated viruses in pDCs, which express preferentially TLR7 and TLR9 [19, 20, 21] . The ''classical'' cytosolic pathway is the major IFN-ab producing mechanism in cells other than pDCs [22, 23] . Signal transduction leading to type I IFN gene induction is initiated by the recognition of intracellular virus-associated molecular patterns. dsRNA and 59-triphosphate RNA produced during viral replication are sensed by the RNA helicases RIG-I and MDA-5 [23, 24, 25, 26, 27, 28] . This pathway has been extensively studied, mainly in virus-infected fibroblasts. Triggering of the aforementioned RNA helicases leads to the activation of the transcription factors NF-kB, IRF-3 and IRF-7 that are important for the induction of IFN-ab and proinflammatory cytokines, including IL-6. In the cytosolic pathway, type I IFN gene induction is a sequential event and both IRF-3 and IRF-7 were shown to be important in the early phase when mostly IFN-b is produced. The late phase of the IFN-ab response is regulated by positive feedback via the increased levels of IRF-7 elicited by IFN-b production during the early phase [29, 30, 31] . In addition to fibroblasts, the potential of mDCs and macrophages to produce significant amounts of type I IFNs in response to viral replication has been demonstrated in vitro [32, 33] ; however, in vivo the specific contribution of these cells to systemic levels of IFN-ab is not well documented. Recently, detection of bacterial DNA in cells infected with L. monocytogenes and recognition of transfected B-DNA has been shown to trigger IFN-b production. This type of response strictly requires IRF-3 [15, 34, 35, 36, 37, 38] . Such a sensing system has been suggested to represent a further mechanism of cytosolic DNA virus recognition [14, 15] and the Z-DNA binding protein 1 (Zbp1, also referred as DNA-dependent activator of IFN regulatory factors, DAI) was shown to be a candidate DNA sensor in this pathway [39] . Notably however, in a follow-up study the same group found a critical role for DAI in L-929 cells but not in mouse embryonic fibroblasts (MEFs) [40] . Furthermore, recent experiments with Zbp1/DAI knockout mice did not show the essential role of Zbp1/DAI in the induction of innate and adaptive responses to B-DNA in vivo and in macrophages, dendritic cells and MEFs in vitro [41] . The induction of type I IFN in Ad-infected mice has been recently studied [10, 42] and associated with both the extracytoplasmic and intracytoplasmic pathways. It was claimed that a part of the IFN-ab response is initiated by TLR9 and MyD88 signaling in pDCs and another part by cytosolic DNA recognition in non-pDCs. However, the identification of IFN-ab producing cell types directly in infected mice was not carried out. In the present study we investigated the IFN-ab responses of Ad-infected mice and showed that the bulk of the in vivo induced IFN-ab is produced by splenic mDCs. Furthermore, we found that TLRs, including TLR9 play no major role. The Ad-elicited IFN-ab response required viral endosomal escape, suggesting a cytosolic induction pathway. Surprisingly however, the induction was independent of IRF-3 and dependent on stress-activated protein kinase/c-Jun NH2-terminal kinase (SAPK/JNK) activity, which is in contrast to the known induction mechanism initiated through cytosolic DNA recognition. Instead, the induction required IRF-7, and a positive feedback regulation via the type I IFN receptor. Although this does not exclude a role for cytosolic nucleic acid sensors, our data do not support the involvement of MDA-5, RIG-I and Zbp1/DAI in the induction of the IFN-ab response to Ad. Furthermore, our results reveal distinct mechanisms in the induction of IFN-ab and IL-6 by Ad. Finally, we show that Ad-induced IFN-ab is a key mediator of hypersensitivity to bacterial lipopolysaccharides in infected mice. Enhanced susceptibility to LPS and to other microbial inducers of inflammation may contribute to toxic reactions observed during adenoviral gene therapy and to the clinical symptoms of adenoviral diseases. In order to characterize the induction of type I IFNs by Ad in vivo, we first examined how two types of human Ad, Ad3 (species B) and Ad R700 (species C) [2] known to use distinct infectious entry routes, elicit an IFN-ab response in vivo. The results summarized in Fig. 1A and supplementary Fig. S1A show that all mice infected with either of the two viruses exhibited similar IFN-ab responses. IFN-ab was first detectable in plasma at 4 h, peaking at 8 h and declining to low levels 18 h after infection. We then investigated, whether the expression of viral genes is required and/or regulate the induction of IFN-ab by Ads. To this end, we injected mice with an UV-inactivated Ad3 (Fig. S1B ) or Ad R700, incapable of viral gene expression, or with a recombinant Ad5 (species C) that contains a deletion of the E1 and E3 Ad early regions and expresses GFP (Ad5-GFP) (Fig. S1C ). As shown in Fig. 1A , UV-inactivated Ads also induced a strong IFN-ab response which, however, peaked at 6 h after injection, i.e. 2 h earlier than the response to intact Ads. A similar early-peaking IFN-ab response was obtained in mice injected with UV inactivated Ad R700 (Fig. S2A ) and with the recombinant Ad5-GFP (Fig. S2B ). Adenoviruses (Ads) are important pathogens and promising vectors for gene therapy applications. In the course of adenoviral infections innate immune responses are activated, which can be beneficial for the antiviral host defense but also detrimental if activated in a deregulated manner. Type I IFNs are crucial for the innate immune control of various viral infections in the mammalian host. So far, the early, systemic release of IFN-ab during viral infections has been attributed to specialized immune cells, the plasmacytoid dendritic cells. Here, in a mouse infection model, we show that wild type Ads, as well as adenoviral vectors, elicit rapid IFN-ab production almost exclusively in another cell population, the splenic myeloid dendritic cells. This IFN-ab storm depends on viral escape from endosomes to the cytosol and the requirements of the response are suggestive of a novel viral induction pathway. Furthermore, we show that virus induced IFN-ab is the key mediator of Ad-induced hypersensitivity to the cytokine-inducing and toxic activity of lipopolysaccharide, a common constituent of Gram-negative bacteria. Since these bacteria comprise several commensals and pathogens, enhanced susceptibility to lipopolysaccharide may contribute to toxic reactions observed during adenoviral gene therapy and to the clinical symptoms of adenoviral diseases. Titration of the viral preparations in mice revealed that a positive correlation between the viral dose and the height of the IFN-ab response existed only when relatively small doses of intact Ads were used. Higher doses of the Ads either did not elicit a further increase of the IFN-ab response (Ad3, Fig. 1B ) or led to a decrease of the response (Ad R700, Fig. S2C ). In contrast, injection of the corresponding UV-inactivated Ads always led to a gradual increase of the IFN-ab response and, at higher viral concentrations, even exceeded the response obtained with intact viruses. Thus, the expression of Ad genes is not required for the induction of IFN-ab in mice. The data also indicate that the expression of early adenoviral genes negatively regulates type I IFN production. Notably, the Ad E1A gene has been shown previously to suppress Newcastle Disease Virus and IRF-3 induced IFN-a4 promoter induction in transient expression assays in fibroblasts [43] . Interestingly, however, viral gene expression did not inhibit the production of IL-6 in either Ad3-or Ad R700-infected mice ( Fig. 1C and Fig. S2D ). Further experiments revealed that, in contrast to the UV-inactivated Ad, heat-inactivated Ad did not elicit IFN-ab in mice (data not shown). Since heat inactivation prevents the entry of Ad into cells ( [44] and Fig. S2E ), we conclude that signal transduction leading to IFN-ab production is activated during Ad entry. Previous studies have shown that, depending on the inducing virus, IFN-ab is produced ubiquitously or in a cell type specific manner [13, 14] . Here we stimulated different primary mouse cells, including bone marrow derived mDCs (BMDC), bone marrow derived macrophages (BMM), bone marrow derived pDCs and mouse embryonic fibroblasts (MEFs), with Ad2 (species C) or Ad3 (species B). Six hours later, the IFN-ab content of culture supernatants and the expression of the viral E1A gene in the cells were determined. BMDC and BMM, but not MEFs, produced IFN-ab (Fig. 1D ), although all cell types were successfully infected as shown by RT-PCR ( Fig. 1E and not shown). Very similar results were obtained when, instead of Ad2, Ad3 or the recombinant Ad5-GFP were used for infection of the three cell types (not shown). In addition, pDCs also produced IFN-ab in response to Ad infection in vitro (Fig. S3A) ; however, only at high multiplicities of infection. Finally, like MEFs, L-929 cells infected with Ad2 produced no IFNab either (Fig. S3B ). In agreement with previous studies [36, 37] , MEFs and L-929 cells produced IFN-ab following transfection with purified DNA (Fig. 1D and Fig. S10C ). In addition to mouse DCs and macrophages we also found that human monocyte-derived DCs produced IFN-ab upon infection with the adenoviral vector Ad5-GFP (Fig. S3C) . The present data confirm that Ads can trigger IFN-ab production in various immune cells in vitro [10, 45, 46] . However, it furthermore indicates that production does not proceed ubiquitously in all types of infected cells. In order to identify the organ site of viral uptake and IFN-ab production in vivo, we analyzed the expression of the early viral gene E1A and of type I IFN mRNAs, respectively, in the spleen, liver, lung and kidney of Ad3-infected mice. We found expression of the early viral E1A gene in spleen and liver (Fig. S1B ), but not in lung and kidney (not shown), which agrees with earlier findings on in vivo Ad tropism [47, 48] . Expression of IFN-a and IFN-b mRNA was below the level of detection in the organs of non-infected controls. We also found that between 4 and 18 h after infection IFN-a and IFN-b mRNAs were expressed at high levels in the spleen (Fig. 1F ), but surprisingly not in the liver, despite the presence of Ad in both organs. As expected, IFN-ab was not expressed in the virus-free lung or kidney of infected mice (Fig. 1F) . In order to identify the cell type(s) producing IFN-ab in vivo, we isolated splenocytes from Ad infected animals 8 h after virus treatment and separated them into CD11c + (DC-containing) and CD11c 2 (non-DC) populations. Both CD11c + and CD11c 2 populations contained viral DNA (not shown). This finding is in accordance with the report of Morelli et al describing that both splenic DC and non-DC contain the virus in Ad-infected mice [49] . However, quantitative RT-PCR determination of IFN-a and IFN-b mRNA in both populations revealed the presence of IFNab mRNAs predominantly in the DC-containing CD11c + fraction ( Fig. 2A, B ), but not in the macrophage containing CD11c 2 fraction. In contrast to IFN-a and IFN-b, the mRNA levels of IL-6, another cytokine known to be induced by Ad, were comparably upregulated in both the CD11c + and CD11c 2 population (Fig. 2C ). Since the latter non-DC population comprises more than 95% of all mouse splenocytes [50] , we conclude that most of the IL-6 made in the Ad-infected spleen is of non-DC origin. According to published data, the CD11c + cell population contains different types of DCs, as well as other cells such as lymphocytes, macrophages, granulocytes and NK cells [51, 52, 53] . We therefore further separated the purified CD11c + cells into mDCs (CD11c + CD11b + Gr1 2 ), pDCs (CD11c + CD11b 2 GR1 + B220 + ) and a CD11c + CD11b-F4/80 + subpopulation ( Fig. S4A -D) and measured the expression of IFN-a, IFN-b and b-actin with real-time RT-PCR. After normalization to b-actin expression, we found that on a per cell basis mDCs expressed significantly more IFN-b than pDCs (Fig. 2D ), but both mDCs and pDCs expressed similar amounts of IFN-a (Fig. 2E ). In contrast, CD11c + , CD11b 2 cells carrying the macrophage marker F4/80 + did not express detectable amounts of IFN-a or IFN-b. Since mDCs comprise approximately 60% of all analyzed CD11c + splenocytes and their numbers are approximately 10-times higher than those of pDCs (Fig. S5A , B and [50] ), these results suggested that the vast majority of IFN-ab in Ad-infected mice was produced by splenic mDCs. To verify this assumption, we analyzed the Ad-elicited cytokine responses in mice depleted of CD11c high MHC II + myeloid DCs. To ablate these cells, we injected diphtheria toxin into the CD11c-diphtheria toxin receptor CD11cDTR/GFP transgenic mice [54] 24 h prior to infection with Ad. In agreement with previous reports [55, 56] , DT pre-treatment of DTR/GFP transgenic mice resulted in the ablation of CD11c high MHC II + CD11b + splenic mDCs, whereas the CD11c int Siglec H + CD11b 2 plasmacytoid DCs remained unaffected (Fig. S5A, B) . When DT pre-treated CD11cDTR/GFP transgenic mice were challenged with Ad3 and examined for IFN-ab in plasma 4 and 8 h after infection, only marginal IFN responses were found at both timepoints, in contrast to the strong responses of similarly infected transgenic control mice that had not received DT (Fig. 3A) . The same pretreatment with DT had no effect on the Ad elicited IFN-ab response of wild-type mice (Fig. 3A) . Interestingly, the determination of IL-6 levels in plasma 8 h after infection revealed that the ablation of mDCs in CD11cDTR/GFP transgenic mice affected only moderately the induction of IL-6 ( Fig. 3B) , confirming that non-DCs contribute significantly to the Ad-elicited IL-6 response in vivo. The fact that different cell types are responsible for IFN-ab and IL-6 response to Ad may explain at least in part why viral gene expression did not inhibit the production of IL-6 in Ad3-or Ad R700-infected mice (see Fig. 1C and Fig. S2D ). In order to functionally evaluate the possible participation of pDCs to the systemic production of IFNab to Ad we also checked this response in mice depleted of pDCs. As shown in Fig. S6A and B, the injection of anti PDCA-1 antibody led to the substantial decrease of the number of splenic pDCs. Nevertheless, the production of IFN-ab was not changed in response to Ad infection (Fig. 3C) . The data collectively show that the vast majority of IFN-ab but not of IL-6 in Ad infected-mice is produced by splenic mDCs. Recent studies have shown the involvement of different TLRs including TLR 2, 3, 4, 7 and 9 in the innate recognition of different viruses [16, 17, 18, 57, 58, 59, 60, 61] . Signaling via TLR9 was shown to be responsible for the strong type I IFN response of pDCs to Ad in vitro [10, 45, 46 ]. Here we investigated the possible involvement of TLR9 in the induction of type I IFNs by measuring IFN-ab in Ad infected wt and TLR9 2/2 mice. Fig. 4A shows that TLR9 2/2 mice produced normal levels of IFN-ab. upon infection with Ad3. Comparable results were obtained with the recombinant Ad5-GFP (Fig. S7A) . Moreover, Unc93B mice, deficient in signalling by intracellular TLRs showed normal IFN-ab responses to Ad5-GFP (Fig. S7A ). We further checked the possible involvement of the TLR system using mice deficient for TLR2, TLR4 or for the TLR adaptor proteins TRIF and MyD88. Fig. 4A shows that the Ad3 induced IFN-ab levels in all strains of mice were as high as in the respective wt controls. Similarly, comparable IFN-ab responses were also found in TLR-, MyD88-or TRIFdeficient mice and in the corresponding wt animals after infection with UV-inactivated Ad3, AdR700 and Ad5-GFP (not shown). Furthermore, comparable IFN-ab responses were found in cultures of Ad-infected BMDCs from wt, MyD88-and TRIFdeficient mice (Fig. 4B ). The various TLR deficient mice showed impaired responses to the corresponding ligands in control experiments (Fig. S7B ). Collectively, our data show that the TLR system plays no major role in the systemic IFN-ab responses to Ads in mice. The type of virus, the IFN-ab-producing target cell, and the activation mechanism determines whether positive feedback signaling is involved in the induction of the IFN-ab response or not [13, 20] . Here, we studied the possible involvement of IFN feedback signaling in the IFN-ab and IL-6 response to Ad3 by using wt and IFN-abR-deficient mice. The absence of the IFN-ab receptor resulted in dramatically decreased levels of IFN-a and IFN-b protein in the plasma as well as of IFN-a and IFN-b mRNAs in the spleen (Fig. 5A , B) 4 and 8 h after Ad infection. The difference between the protein or mRNA levels of IFN-a and IFN-b in wt versus mutant animals was approximately 100 and 20fold, respectively. Furthermore, in contrast to wt mice, the characteristic rise in IFN-ab levels between 4 and 8 h after infection was absent in IFN-abR 2/2 mice. Thus, in Ad-infected mice, production of both IFN-a and IFN-b is strictly dependent on positive IFN-ab feedback. The determination of IL-6 protein and mRNA levels in the same plasma and splenic tissue samples revealed that the induction of IL-6 is also positively regulated by IFN-ab signaling (Fig. 5C ), which is in agreement with a previous study [10] . We also tested whether IFN-abR-dependent signaling is involved in the IFN-ab response of Ad-infected BMDCs in vitro. As with the in vivo results, we found that cells from IFN-abR 2/2 mice infected with Ad3 produced significantly less type I IFN than similarly infected cells from wt mice (Fig. S8 ). The loss of IFN-ab signaling also resulted in a strong inhibition of the Ad induced IL-6 production in BMDCs (Fig. 5D ) and BMMs (Fig. 5E ). Furthermore, as shown using Ad-infected BMMs, it resulted also in a strong reduction of inducibility of IL-6 mRNA expression (Fig. 5F ). Because transcriptional changes are often determined by epigenetic factors [62] we checked the levels of hyperacetylated histone H4 (acH4), a permissive factor for transcription at the IL-6 promoter in control and Ad-infected BMMs from wt and IFNabR 2/2 mice. Chromatin immunoprecipitation (ChIP) assays showed a significant enrichment of acH4 at the IL-6 promoter in infected wt, but not IFNabR 2/2 BMMs (Fig. 5G) . In a control experiment, as expected, an enrichment of acH4 was observed at the promoter of the constitutively active Topoisomerase 3b but not of the l5 (not expressing, active only in early B-cells) gene in both cells types. From these results we conclude that IFN-ab exerts a positive regulatory effect on the Ad-induced IL-6 transcription and that its loss is at least partly responsible for the strong reduction of the IL-6 response in Ad-infected IFN-abR knockout mice. Using real-time RT-PCR we then analyzed the spectrum of IFN-ab genes in wt mice as well as the impact of IFNabR deficiency on their induction. Included were IFN-a2, 4, 5, 6, 9, 11, 12, 13, 14 and IFN-b. All of them were induced in the spleen by Ad3 in vivo and in BMDC in vitro. IFN-ab subtypes were not detectable in the spleen of unstimulated mice or BMDCs (not shown). Fig. 6 shows the patterns of IFN-ab genes induced in vivo and in vitro in the presence or in the absence of IFN-ab feedback signaling. In wt mice and cells IFN-a5 and IFN-b were the most strongly expressed genes and IFN-a13 was the least activated IFNa subtype. IFN-a11 was not induced at all. IFNabR deficiency resulted in an inhibition of IFN-ab gene expression, the strength of which was subtype dependent. In some cases the inhibition was weak (IFN-a2, 4, 5 and IFN-b), in others strong or complete (IFN-a12, 13, and 14) , showing that the expression of different subtypes of IFN-ab are differentially affected by IFN-ab feedback signaling. Collectively, these data show that the adenovirus triggered production of IFN-ab and IL-6 in BMDCs in vitro and in mice in vivo is strongly dependent on intact IFN-ab signaling. The critical role of IRF-7 but not IRF-3 in Ad-induced type I IFN production The transcription factors IRF-3 and IRF-7 have distinct and important roles in IFN-ab production induced by viruses or other pathogens and their involvement can be characteristic for the induction mechanisms involved [13] . Specifically, IRF-3 has been shown to be critically involved in cytoplasmic DNA sensing and in the Ad-induced IFN-ab production in BMMs in vitro [63] . We show here that IRF-3 is critical for the induction of IFN-ab by isolated adenoviral DNA, but not by infection with whole virions in BMDCs (Fig. 7A ). In order to analyze the individual contribution of IRF-3 and IRF-7 to the Ad-induced IFN-ab response in vivo, we infected mice deficient for these transcription factors with Ad3. We also compared these responses to those triggered by poly I:C, a known activator of the cytosolic IFN-ab producing pathway in vivo. As shown in Fig. 7B , the lack of IRF-3 did not significantly influence the plasma levels of IFNab 4 or 8 h after infection in response to Ad. In contrast, Ad-infected IRF-7deficient mice did not exhibit detectable amounts of IFN-ab in the plasma. Very similar data were obtained with Ad5GFP (Fig. S9) . Thus, IRF-7, but not IRF-3, is essential for the induction of the IFN-ab response during Ad infection. Compared to wt mice, poly I:C injected IRF-7 deficient mice produced significantly less, but still well detectable amounts of IFN-ab. The nucleic acid sensors MDA-5, RIG-I and DAI/Zbp1 are not critical for Ad-induced type I IFN production Next, we investigated whether the cytoplasmic RNA sensors MDA-5 or RIG-I, or the putative DNA sensor DAI/Zbp1 may play a major role in the induction of the IFN-ab response to Ad. The possible involvement of MDA-5 was tested using BMDCs and BMMs from MDA-5 deficient mice. Normal IFN-ab responses to Ad were obtained in MDA-5-deficient BMDCs cells (Fig. 8A ) and also in BMMs (not shown), whereas the responses to the known MDA-5 ligand Poly I:C were abrogated (Fig. 8A) . The possible involvement of RIG-I was checked using a dominant negative form of RIG-I (RIG-IC) [64] . BMDCs from IRF-3 2/2 mice were transfected with a GFP-expressing plasmid (transfection control) with or without RIG-IC and subsequently stimulated with Ad or control leader RNA. We used IRF-3 2/2 cells to avoid any induction of IFN-ab by the plasmid itself, which is, contrary to that by Ad, strictly IRF-3 dependent. The induction of IFN-b mRNA was analyzed in sorted GFP-positive cells. As shown in Fig. 8B , left, the transfection of RIG-IC prevented induction of IFN-b mRNA by the leader RNA, but not by Ad. The role of Zbp1/DAI was analyzed using siRNA-mediated knockdown of DAI/Zbp1 in BMDCs. For this purpose, cells from IRF-32/2 mice were co-transfected with DAI/Zbp1 targeting siRNAs and a GFP expressing plasmid and subsequently stimulated with Ad. As shown in Fig. 8B , right, the transfection of BMDCs resulted in strongly reduced DAI/Zbp1 expression but not in a reduced IFNb mRNA induction by Ad in GFP-positive cells. Control experiments with L-929 cells showed that transfection of the gene-specific siRNA downregulated the expression of DAI/Zbp1 on both the mRNA and protein levels and efficiently inhibited the IFN-b response to B-DNA ( Fig. S10A-C) . Collectively, our results indicate that known nucleic acid sensors such as MDA-5, RIG-I and Zbp1/DAI are not involved in the Ad-induced type I IFN production. MAPKs have been previously shown to be activated by Ad in vitro, in different non-immune cell types and to be important for the induction of chemokines in response to Ad [65, 66, 67, 68] . Here we investigated whether members of the MAPK family play a role in the Ad-induced IFN-ab and IL-6 response. We infected BMDCs with Ad3 in the presence or absence of MAPK inhibitors and determined the levels of IFN-ab and IL-6 produced. Fig. 9A and Fig. S11 show that the pharmacological blockade of the SAPK/JNK MAPK almost completely inhibited the Ad3-induced production of both IFN-ab and IL-6. In contrast, the inhibition of the p38 MAPK pathway partially inhibited the production of IL-6, but had no effect on the production of IFN-ab. Finally, the blockade of ERK1/2 had no effect on the production of either IL-6 or IFN-ab. Very similar data on the effects of MAPK inhibitors were obtained using Ad2 and mutant Ad5-GFP to stimulate BMDC (data not shown). Next, we analyzed the levels of activated SAPK/JNK MAPK proteins in BMDCs and found their robust phosphorylation 2 h after either Ad3 or Ad5-GFP infection (Fig. 9B) . We also tested the importance of SAPK/JNK signaling on Ad-induced IFN-ab production in vivo. Fig. 9C shows that the blockade of the SAPK/ JNK signaling pathway in mice completely inhibited the production of IFN-ab at 4 and partially at 8 h after Ad infection. Taken together, these data strongly indicate that the Ad-activated SAPK/JNK MAPK pathway plays an important role in the virusinduced production of type I IFNs and IL-6. Endosomal escape triggers the Ad-induced production of IFN-ab and IL-6, but prevents TLR9-dependent innate recognition During the course of our adenovirus preparations, we regularly found ''empty capsids'' which we separated and purified in addition to the mature virions. We tested the IFN-ab stimulating activity of these preparations in vivo and observed that they were not active (Fig. S12A ). Since empty capsids lack viral DNA and exhibit an altered protein composition [69] , the absence of an inducing viral constituent(s) from these capsids could explain their inability to provoke an IFN-ab response. Another possible explanation could be that endosomal escape is required for IFNab induction, since empty capsids cannot escape from the endosome [69] . To test the latter possibility, we infected mice with 3.6610 10 viral particles of wt Ad2 and Ad2Ts1, a viral mutant deficient in endosomal escape [5] . As shown in Fig. 10A , in contrast to the wt virus, Ad2Ts1 did not induce detectable levels of type I IFN at 4 and 8 h after infection. Furthermore, the IL-6 response was also severely reduced in these animals (Fig. S12B) . In a control experiment, mice infected with Ad2Ts1 grown at permissive temperature (32uC) and thus capable of endosomal escape, exhibited normal IFN-ab responses (Fig. 10B) . The inability of Ad2Ts1 to escape from endosomes of mDCs was confirmed by electron microscopy (Fig. 10E-H) and in vivo by the lack of Ad early E1A gene expression in the spleen of mice infected with Ad2Ts1 (Fig. 10I) . Thus, escape from the endosome is critical for the induction of IFN-ab and IL-6 by adenoviruses. It should be noted that a further increase in the Ad2Ts1 dose (2.16610 11 particles) resulted in detectable, albeit very low levels of plasma IFN-ab that were released with different kinetics (Fig. 10C) . In this case, IFN-ab was detectable as early as 2 h after infection and, in contrast to the results we obtained with wt viruses ( Figures 1A, 10A and S1A), the levels of IFN-ab did not increase significantly at the later time-points. This already suggests a mechanism for type I IFN induction by Ad2Ts1 that is fundamentally different from the IFN-ab induction seen with wt Ad2. Since Ads are DNA viruses, they can possibly be detected by TLR9. In fact, the innate immune recognition of Ad in pDCs is TLR9 dependent. We therefore repeated the experiment with a high dose of Ad2Ts1 using TLR9 2/2 mice. As shown in Fig. 10C , TLR9 2/2 mice did not produce IFN-ab in response to the mutant Ad2Ts1, quite in contrast to the results obtained with the wt virus (see Fig. 4A ) Likewise, there was no detectable IFN-ab release in Ad2Ts1-infected mice deficient in MyD88, an essential component of TLR9 signaling (not shown). Furthermore, the corresponding amount of empty particles (DNA-free) of Ad2Ts1 elicited no IFN-ab response in wt mice (Fig. 10C) . These data illustrate the critical role of TLR9 in the induction of IFN-ab by means of high doses of Ad2Ts1. To exclude the possibility that contaminating DNA on the surface of the virions was responsible for the TLR9dependent IFN-ab induction, we treated the mutant virions with bensonase, which destroys all kinds of free nucleic acids. Bensonase-treated Ad2Ts1 still induced an IFN-ab response in We also tested the role of endosomal escape of Ad in the in vitro induction in BMDCs of IFN-ab and IL-6. Fig. 10D shows a dose dependent production of IFN-ab by Ad2Ts1 grown at permissive temperature and Figures S12C and D the production of IFN-ab and IL-6 by wt Ad2, but no production of either of the cytokines by Ad2Ts1 grown at the restrictive temperature. Interestingly, liposomal transfection of whole Ad2 Ts1 virions (but not of empty virus particles) in BMDCs resulted in the significant production of IFN-ab that however, was critically dependent on IRF-3 (Fig. 7A) . Similarly, Ad2 Ts1 did not induce IFN-ab production in human monocyte derived DCs (Fig. S12E) . The requirement of low pH for Ad3 infection was also tested using bafilomycin A1, a drug known to inhibit the acidification of endosomes. Experiments shown in Fig. S13A , B revealed that cells treated with this drug produced significantly reduced levels of IFN-ab and IL-6 respectively, in response to Ad3. This is consistent with the notion that Ad3 and Ad7 infection of cultured cells requires low endosomal pH [70, 71] . Similar results were obtained using treatment with ammonium chloride, another acidification inhibitor known to block Ad escape from endosomes (not shown). Collectively, the data in vitro and in vivo provide evidence that the late phase of the Ad infectious entry, in which the virus escapes from the endosome, triggers an innate response characterized by the production of type I IFNs and IL-6. Induction of type I IFNs was shown to be critical for some innate immune responses to Ads [10] .We therefore investigated whether a characteristic consequence of infection with Ad, the induction of LPS hypersensitivity [72] , might be mediated by type I IFNs. For this purpose, we infected wt and IFNabR 2/2 mice with Ads, challenged them 16 h later with LPS and measured the TNF-a response. Non-infected LPS-treated mice served as a control. Unlike the infected wt mice, the Ad3-infected IFNabR 2/2 mice did not exhibit enhanced responses to LPS (Fig. 11A) . Interestingly, LPS hypersensitivity developed also in mice injected with very small amounts of Ad. Such amounts were capable of the elicitation of IFN-b mRNA in the spleen of infected animals, but incapable of inducing detectable circulating IFN-ab ( Fig. S14A and B) . Similar results were obtained in Ad5-GFP-infected mice (not shown). Furthermore, in IRF-7 2/2 mice that exhibit a severe impairment of the IFN-ab response to Ad a severe impairment of LPS sensitization by Ad5-GFP was also observed. (Fig. 11B) . In contrast, sensitization to LPS developed normally in Ad5-GFPinfected TLR9 2/2 mice (Fig. 11B) , which is in agreement with our finding that TLR9 is not critical for the induction of IFN-ab by Ad. Furthermore, only Ad5-GFP-infected wt, but not IFNabR 2/2 mice exhibited enhanced susceptibility to LPS shock (Fig. 11C) . Notably, also Ad vector-infected or IFN-a treated human monocyte-derived mDCs exhibited enhanced sensitivity to LPS and overproduced TNF-a upon LPS challenge (Fig. S14C) . On the whole, our results indicate that IFN-ab is an essential mediator of the LPS hypersensitivity induced by adenovirus infection. An increased expression of the LPS receptor complex on target cells, may contribute to the enhanced reactivity to LPS [73, 74] . We therefore investigated whether macrophages of Ad infected mice overexpress the receptor components mCD14 and TLR4/ MD-2. We found that splenic macrophages from Ad infected mice overexpress mCD14 in an IFN-ab dependent manner (Fig. 11D ) but the expression of TLR4/MD-2 was only minimally affected (Fig. S14D) . Furthermore, we found increased acetylation levels of histone H4 at the TNF-a promoter in Ad-infected wt, but not in IFNabR-deficient macrophages (Fig. 11E ). This suggests that Adinduced IFN-ab increases LPS-induced TNF-a production, at least in part by epigenetic changes at the TNF-a promoter. In the present study we investigated the IFN-ab response of mice infected with human Ads. We showed that the response is characterized by high levels of IFN-a and IFN-b, which are produced simultaneously and almost exclusively by splenic mDCs. Furthermore, the response is entirely independent of viral replication, TLR-, MyD88-, TRIF-and IRF-3-signaling, but dependent on viral escape from the endosomes, activation of the SAPK/JNK pathway and IRF-7, and subsequent IFN-ab feedback signaling. These data suggest an IFN-ab induction pathway that is different from the extracytoplasmic and cytoplasmic pathways so far described. Finally, we show that IFN-ab is a key mediator of the hypersensitivity to bacterial lipopolysaccharide, which develops, in Ad-infected mice. As shown previously and in this study, a wide spectrum of cells including pDCs, mDCs and macrophages [10, 45, 46, 63] produce IFN-ab in response to Ad in vitro. The present finding that in vivo, in Ad-infected mice, splenic mDCs are the major source of IFN-ab is surprising, since in mice infected with different viruses (MCMV, HSV, VSV, MHV, influenza, vaccinia and Sendai viruses) [27, 57, 58, 59, 75, 76, 77, 78] pDCs activated by the TLRs constitute the major contributors to the systemic levels of type I IFNs. In the present study the expression levels of IFN-ab mRNAs in organs and cells from Ad-infected mice suggested a dominant role for splenic mDCs in the IFN-ab response. Furthermore, the IFN response to Ad was practically absent in mice depleted of CD11c high MHC II + myeloid DCs. Also, there was a striking similarity between the IFN-ab subtypes induced by Ads in the spleen of infected mice and those induced in mDC cultures in vitro. It should be emphasized that in mice infected with VSV, in which splenic pDCs are the main IFN producers, the spectrum of in vivo induced IFN-a subtypes was markedly different [76] . Finally, the finding that the response of Ad-infected mice was completely independent of TLR signaling and strongly dependent on IFN-ab feedback provide further arguments against a major role for pDCs. In a number of viral infection models, IFN-a and -b production by pDCs was mediated by TLR7/9 [13, 19, 23] and was at least partially independent of a positive IFN-ab feedback [76, 77, 79, 80] . In variance to our data, pDCs and various types of non-pDCs [10, 42] were suggested to be responsible for the type I IFN responses to adenoviral vectors in mice. In [10] the loss of TLR9 signaling resulted in a reduction of the IFN-a response in mice infected with the recombinant species C Ad-lacZ [10] . This finding suggested a significant contribution of TLR9, and therefore of pDCs, to the Ad-induced IFN-ab response in vivo. The ratio of pDCs to mDCs (approximately 1:1) in the spleen of animals used in the above study was quite different from that we and others [50] have found (approximately 1:10). This may explain the conflicting results on the role of TLR9 in vivo, between this previous study and ours. In another study [42] the induction of IFN-ab was studied in mice with an artificially enlarged pool of DCs (due to prior pretreatment with sFLT-3L), 24 h after recombinant Ad administration. In our study naïve mice were used and 24 h after Ad infection the levels of IFN-ab were already below the detection limit. Of note also, the bone marrow stromal antigen 2 (BST2) (that is recognized by the PDCA-1 antibody used for isolation of pDCs in [42] ), was shown to be up-regulated on numerous cell types following stimulation that triggers an IFN response [81] . Thus, the use of the PDCA-1 antibody for the isolation of pDCs seems to be more reliable in the case of uninfected mice [81] . Recently, an absolute IRF-3 dependency of the in vitro IFN-ab response of bone marrow derived macrophages to Ad has been reported [63] . In our study IRF-3 deficiency had no significant effect on the levels of IFN-ab induced in Ad-infected mice. In addition, in Ad-infected macrophages, the relatively low level of IFN-a4 mRNA and its negative regulation by the autocrine Figure 11 . Adenovirus triggered IFN-ab production is the mediator of LPS hypersensitivity. Wild-type and IFN-abR 2/2 mice (4-6 mice/ group) were infected i. p. with 1610 10 Ad3 particles or left uninfected. 16 h later the animals received 1 mg LPS i. p. or diluent only. TNF-a was determined in plasma 2 h after challenge (A). One representative experiment of three is shown. n.d.: not detectable. IRF-7 2/2 and TLR9 2/2 mice were infected i. p. with 2610 9 Ad5 GFP particles and injected 16 h later with 1 mg LPS i. p. TNF-a was determined in plasma 2 h later (B). IFN-ab signaling is required for Ad infection augmented LPS lethality. Wt and IFNabR 2/2 mice were infected with 1610 10 Ad5 GFP particles i. p. or left uninfected. 16 h later mice were injected with 3.5 mg LPS/gr. bw. The number of dead/total animals are shown at the indicated time after LPS challenge (C). Ad infection increases mCD14 expression on splenic macrophages in an IFN-ab signaling dependent manner. Wt and IFN-abR 2/2 mice were infected with 1610 10 Ad5 GFP particles i. p. or left uninfected. Expression of mCD14 was measured on the surface of F4/80 + splenic macrophages by FACS 16 h after infection (D). Ad infection increases acetylation of histone H4 in wt macrophages at the TNF promoter. Wt and IFNabR 2/2 BMMs were mock-infected or infected with 5400 Ad5 GFP particles/cell and levels of acetylation of histone H4 were measured at the TNF-a promoter with ChIP assays (E). Representative experiments of two are shown. doi:10.1371/journal.ppat.1000208.g011 feedback was different from our in vivo findings (strong induction and positive regulation by feedback). This is in agreement with the concept that in vivo macrophages make no significant contribution to the IFN-ab response to Ad. A likely explanation is that the induction of IFN-ab in macrophages requires the high numbers of virions used in vitro, and that such multiplicities were never reached in vivo. Interestingly however, in contrast to IFN-ab induction, the induction of IL-6 proceeded in the spleen of infected mice in both DC and non-DC populations. In accordance, the depletion of IFN-ab-producing mDCs in mice prior to infection lowered, but did not entirely prevent the IL-6 response. It is possible that the pathways leading to the induction of IFN-ab and IL-6 by Ad are different, at least in different cell types. Alternatively, in vivo, part of the IL-6 formed is induced indirectly via secondary mechanisms. In this context it is interesting that blockade of the p38 MAPK in mDCs in vitro had no effect on Ad-induced IFN-ab production, but partially inhibited the production of IL-6. Our data showing that IFNabR 2/2 mice exhibit strongly impaired IL-6 responses to Ad is in agreement with a previous report [10] and shows that IFN-ab is a positive regulator of IL-6 production. Moreover, our data indicate that this effect could be explained at least in part by the positive regulatory effect of IFN-ab on Ad-induced IL-6 transcription and is achieved by the alteration of the chromatin structure at the IL-6 promoter. The experiments carried out in this study in MyD88-, TRIF-, Unc93B and various TLR-deficient mice excluded a participation of TLRs in the IFN-ab responses to Ads, including the recombinant Ad5-GFP. The only exception was the TLR9-and MyD88-dependent IFN-ab response elicited by Ad2Ts1, a mutant virus deficient in endosomal escape [5] . However, compared to all other Ads used in this study, Ad2Ts1 induced very low levels of IFN-ab, with faster kinetics and only when used in very high amounts. This is consistent with the finding that the mechanisms of type I IFN induction by Ad2Ts1 and the other adenoviruses are not the same. We suggest that the fast escape of Ads from the endosome circumvents activation of endosomal TLRs. The requirement for a longer endosomal retention time in TLR9dependent IFN-ab production has been recently demonstrated [82] . On the whole, our experiments indicate that Ad endosomal escape is required for the induction of IFN-ab in vivo and suggest a cytosolic pathway. The same requirement was ascertained in the present study for the IL-6 response. Sensors of nucleic acids are powerful initiators of the TLRindependent cytosolic IFN-ab induction [12, 14, 15] . We excluded in this study a major involvement of the cytoplasmic RNA sensors RIG-I and MDA-5 in the induction of IFN-ab by Ads. Likewise, our study does not support a major participation of the cytosolic DNA sensor DAI/Zbp1 in this induction either. However, our study did not formally exclude the existence of all potentially redundant recognition pathways. So far reported, the pathway(s) of IFN-ab induction activated by cytosolic DNA is (are) strictly dependent on IRF-3 and minimally on IRF-7 [34, 35, 36, 37, 39] . IRF-3 was also reported to be essential for the adenoviral DNAdependent induction of IFN-ab in BMMs in vitro [63] . In agreement, in this study we show that transfection of BMDCs with naked adenoviral DNA or with whole virions of the endosomal escape deficient Ad2Ts1 results in a strictly IRF-3 dependent IFN-ab response. Evidently however, the IFN-ab response of Ad-infected BMDCs and mice is independent of IRF-3. These findings do not exclude a requirement for dsDNA recognition in the induction of IFN-ab, but suggest a different induction mechanism and show the importance of cellular compartmentalization during normal Ad entry. A possible important factor involved in cytosolic Ad sensing might be the adenoviral cysteine protease L3/23, whose activation requires the presence of Ad DNA [83] . This assumption is supported by our finding that a small fraction of the protease-lacking Ad2Ts1 still reaches the cytosol (Fig. 10F, H) , but is devoid of any IFN-abinducing activity. This enzyme, apart from its involvement in maturation of viral proteins and endosomal escape, is essential for the stepwise disassembly of Ads in the cytosol and the release of viral DNA at the nuclear pore [5] . The absolute requirement of IRF-7 for type I IFN induction in Ad-infected mice shows that this transcription factor participates not only in the positive IFN-ab feedback, but also in the initial IFN-ab production and indeed plays a master role in the regulation of type I interferon response [29] . A further finding of the present study is the essential role for the MAPK SAPK/JNK in the IFN-ab induction by Ads in vitro and in vivo, although the DNA-mediated cytosolic induction of IFN-ab has been reported to be independent of MAPKs activation [14, 15, 36] . A cytosolic dsDNA-signaling pathway, mediated by the RNA helicase RIG-I and MAVS, and leading to the induction of IFN-b has been recently demonstrated in human hepatoma cells [84] . This induction pathway is absent in murine systems [22, 37, 84, 85] and is therefore unlikely to participate in the IFN-ab response of Ad-infected mice. On the whole, our findings do not support the participation of any known nucleic acid-mediated mechanisms in the elicitation of IFN-ab responses to human Ads. In this context it is interesting that, as shown here, mouse embryonal fibroblasts do not produce IFN-ab upon Ad infection, although they posses efficient cytosolic induction pathways for dsRNA or dsDNA [13, 36, 37] , the latter shown also in this study. Rather, our data support the possibility that the IFN-ab response to Ads occurs via a novel, not yet characterized cytosolic DNA or protein recognition pathway. Further studies are required to identify the viral components and the host receptors involved. We also emphasize that the known extra-and intra-cytosolic induction pathways may contribute to the IFN-ab response in a host in which Ads replicate and free viral dsRNA and dsDNA are generated. As mentioned above, the production of IFN-ab in Ad-infected mice is strongly dependent on IFN-ab feedback signaling. Inhibition of positive IFN-ab feedback is a likely explanation for the negative regulation of the IFN-ab response, observed, in mice infected with intact Ads (expressing early genes) in this study. Likely candidates for negative regulators of the IFN-ab production are the E1A proteins. Previously, they were shown to inhibit Stat1 signal transduction [86] , which plays a role in positive feedback signaling by means of the IFN-ab receptor. Because the adenoviral vectors used in this study correspond to some of those used in gene therapy trials, the present findings in mice may have important implications for Ad gene therapy applications. The adverse effects observed in therapeutic trials, such as systemic inflammatory response and toxicity [6, 7, 9] can be at least partly explained by the enhanced susceptibility of the Adinfected host to microbial components, such as LPS [72] and lipopeptides (unpublished results) from incoming secondary pathogens or from the patient's own flora. Our in vitro finding of a strongly enhanced TNF-a response to LPS in Ad5-GFP-infected human DCs suggests that enhanced susceptibility to LPS may develop also in patients treated with adenoviral vectors. As shown here, this hypersensitivity is mediated by viral-induced IFN-ab, which is in accordance with the role of IFN-ab as a key mediator of sensitization to LPS [87, 88] . The increased mCD14 expression on LPS target cells and epigenetic changes on promoters of relevant genes (both shown here), may at least in part explain the role of IFN-ab in the development of Ad induced LPS hypersensitivity. Type I interferon induction was recently found in recombinant Ad-treated human cells, especially in pDCs [46, 89] and in monocyte-derived DCs in the present study. Moreover, this response was observed also in patients administered with recombinant adenoviral vectors [42, 90] . Since as shown here, IFN-a pre-treatment is capable of increasing susceptibility to LPS of human DCs in vitro, we assume that Ad-induced IFN-ab can induce LPS hypersensitivity in humans in vivo. Undesirable complications mediated by IFN-ab can occur not only during gene therapy, but also in immunocompromised patients where Ads are major pathogens [1] . Our finding that blocking SAPK/ JNK signaling inhibits the IFN-ab response to Ads, is of potential interest for prevention or treatment of the direct and indirect adverse effects of IFN-ab in Ad gene therapy. Wt C57BL/6, C57BL/10 and 129Sv mice, as well as all knockout mice were bred under SPF conditions at the MPI. Breeding pairs of IRF-3 2/2 and IRF-7 2/2 mice were kindly provided by T. Taniguchi [54] were on a C57BL/6 background, TLR2/4-deficient mice [95] on C57BL/10 background and IFNabR-deficient mice [96] on 129Sv and for the lethality experiments on C57BL/6 background. When the strain of the mouse was not indicated C57BL/6 mice were used. Mice of both sexes, 8-12 weeks of age were used for the experiments. All of the experimental procedures were in accordance with institutional, state and federal guidelines on animal welfare. Human Ads of species B (Ad serotype 3) and C (Ad R700, an Ad serotype 5 derivative, Ad serotype 2, Ad2Ts1 and Ad5-GFP an early gene expression defective Ad were grown, purified and stored as previously described [5, 72] . The ratio of infectious/total viral particles was determined on susceptible cells and was typically 1:20-50. If not otherwise stated, Ad2Ts1 used for the experiments was grown at non-permissive temperature 38.5uC (resulting in the absence of incorporation of the adenoviral protease L3/23 into viral capsids). Empty particles were identified by their light density in CsCl density gradients and the absence of viral DNA and were purified simultaneously with mature virions. UV inactivation of Ad was done as described previously [5, 72] using the minimal essential dose preventing viral gene expression and replication in susceptible cells. Heat inactivation was done at 56 uC for 60 min [44] . All virus preparations were LPS-free (less than 1 pg LPS/ 10 11 viral particles) as determined by the Limulus amebocyte lysate test (Pyroquant Diagnostic GMBH, Mörfelden, Germany). MEFs, BMMs and GM-CSF induced BMDCs were generated as described [95, 97] . The purity of BMMs was higher than 98%, of BMDCs approximately 80%. BM derived pDCs were generated in the presence of Flt3L and CD11c + CD11b 2 B220 + CD62L + pDCs were MoFlo sorted (purity higher than 95%) as described [55] . MEFs were grown in a-MEM (Invitrogen). Immature, monocyte derived human DCs were obtained by incubating adherent monocytes with GM-CSF as described [98] Human mDCs were infected with the indicated amounts of Ads in growth medium containing 2% of donor serum. Mouse L-929 cells were grown in DMEM with 10% FCS. MAPK inhibitors UO126 (MEK1/2, Cell Signaling), SB203580 (p38, Sigma) and SP600125 (SAPK/JNK Calbiochem) were used at 15 mM in vitro and SP600125 at 20 mg/kg in vivo. Bafilomycin-A1 (Sigma) was used at 100 nM. The plasmids pmaxGFP (Amaxa) and RIG-IC [64, 99, 100] were purified with the Endo-Free Plasmid kit (Qiagen) and PEG purification. Murine IFN-ab activity was measured using an L-929 cell line (provided by B. Beutler and Z. Jiang, Scripps Research institute, La Jolla) as described [18] Human IFN-ab bioactivity was measured using the HL116 cells from G. Uzé as described [101] . The contribution of IFN-a or IFN-b to the total IFN-ab activity was determined by pre-incubating plasma for 1 h with excess amounts of neutralizing anti-IFN-b antibody (Yamasa Corporation, Japan) or control antibody. Murine TNF-a was measured by a bioassay as described [95] . IL-6 was detected with an ELISA from Pharmingen BD. Human TNF-a was detected with an ELISA from R&D Biosystems. JNK/SAPK MAPKs were detected on immunoblots with antibodies detecting all or phosphorylated isophorms of the proteins (Cell Signaling). Zbp1/DAI and b-actin were detected on immunoblots with antibodies from Santa Cruz Biotechnology. Conventional RT-PCR, real-time RT-PCR, analysis of IFN-áâ subtypes and Chromatin Immunoprecipitation Total RNA was isolated from organs and cells with guanidinium-thiocyanate-phenol-chloroform extraction or with TRI reagent (Sigma). To exclude DNA contamination, RNA samples were treated with RNase free DNase I (Fermentas). cDNA was prepared using Expand reverse transcriptase (Roche) and oligo-dT. Conventional RT-PCRs were performed with primers as follows. [102, 103] , detecting all IFN-a mRNAs. To determine the levels of different IFN-ab subtypes the HT7900 quantitative PCR system (Applied Biosystems) was used. cDNAs were measured in duplicates or triplicates using the following genespecific assays (TaqMan Gene Expression Assays, Applied Biosystems): IFN-alpha2 (Mm00833961_s1), IFN-alpha4 (Mm00833969_s1), IFN-alpha5 (Mm00833976_s1), IFN-alpha6 (Mm02524285_g1), IFN-alpha9 (Mm00833983_s1), IFN-alpha11 (Mm01257312_s1), IFN-alpha12 (Mm00616656_s1), IFN-al-pha13 (Mm00781548_s1), IFN-alpha14 (Mm01703465_s1),IFNbeta (Mm00439546_s1). The gene for mouse hypoxanthine guanine phosphoribosyl transferase-1 (HPRT-1, Mm00446968_m1) was used to calibrate the mRNA levels. Quantitative analysis was performed using the SDS 2.1 software (Applied Biosystems). mRNA levels were calculated by the following formula: relative expression = 2'(2(Ct(Target)-Ct(Endogenous control))*f, with f = 10 000 as an arbitrary factor. Chromatin Immunoprecipitation (ChIP) assays were performed as described [104] . Briefly, cells were cross-linked with 1% formaldehyde and sonicated nuclear extracts were mock-immunoprecipitated or immunoprecipitated with anti-tetraacetylated histone H4 (Upstate). Recovered DNA aliquots from these samples and from input extracts were amplified with real-time PCR using the LightCycler II system (Roche) and Quantitect SYBR Green PCR Kit (Qiagen). Enrichments at specific chromatin loci are shown as the amount of immunoprecipitated DNA in the percent of total input chromatin. The following primers were used: IL -6 promoter: TGG GGA TGT CTG TAG CTC ATT and CAT AGC GGT TTC TGG AAT TGA, TNF promoter: GGG CAG CCC CAG AGG GAA TGA ACT C and TAT GGC AGA GGC TCC GTG GAA AAC TCA CT, Topoisomerase 3b promoter: AGT CCG AGA ACA GCC TGG GT and AGT TGT GCT GCC CAC AGA GG, l5 promoter: TCC CCA TTG CCA GAT AGA GAC ACA and TGG GCC CAA CAG ATT AAC ACA GAG. BMDCs were cold synchronized with saturating amounts of Ad2 and Ad2-ts1 (60 mg/ml, 0.25 ml per 4610 4 cells on a 12 mm glass coverslip) for 1 h, washed and incubated at 37uC for the indicated times. The samples were fixed with 2.5% glutaraldehyde in 0.1 M ice-cold Na-Cacodylate buffer (pH 7.2) containing 0.5 mg/ml ruthenium red for 1 h, washed with 0.1 M Na-Cacodylate buffer (pH 7.2), post-fixed with 2% OsO4 in the same buffer containing 0.5 mg/ml ruthenium red for 1 h at room temperature, and embedded in Epon as described [105] . Virus particles at the plasma membrane, endosomes and the cytosol were determined, and results expressed as means of analyzed cells (n) with standard errors of the mean. List of accession numbers/ID numbers for proteins mentioned in the text The impact of adenovirus infection on the immunocompromised host The coxsackievirus-adenovirus receptor protein can function as a cellular attachment protein for adenovirus serotypes from subgroups A Adenoviridae: The viruses and their replication A mouse model for investigating the molecular pathogenesis of adenovirus pneumonia The role of the adenovirus protease on virus entry into cells The innate immune response to adenovirus vectors Fatal systemic inflammatory response syndrome in a ornithine transcarbamylase deficient patient following adenoviral gene transfer Cytokines in adenoviral disease in children: association of interleukin-6, interleukin-8, and tumor necrosis factor alpha levels with clinical outcome Interference with the IL-1-signaling pathway improves the toxicity profile of systemically applied adenovirus vectors Innate immune response to adenoviral vectors is mediated by both Toll-like receptor-dependent and -independent pathways Interferons, interferon-like cytokines, and their receptors Type 1 interferons and the virus-host relationship: a lesson in detente Type I interferon [corrected] gene induction by the interferon regulatory factor family of transcription factors Type I interferons in host defense Innate immune recognition of, and regulation by Pattern recognition receptors TLR4 and CD14 mediate response to respiratory syncytial virus Subversion of the innate immune system by a retrovirus CD14 is required for MyD88-independent LPS signaling IPC: professional type 1 interferon-producing cells and plasmacytoid dendritic cell precursors Production of type I interferons: plasmacytoid dendritic cells and beyond Plasmacytoid dendritic cells in immunity The specific and essential role of MAVS in antiviral innate immune responses Cell type-specific involvement of RIG-I in antiviral response Shared and unique functions of the DExD/H-box helicases RIG-I, MDA5, and LGP2 in antiviral innate immunity Essential role of mda-5 in type I IFN responses to polyriboinosinic:polyribocytidylic acid and encephalomyocarditis picornavirus RIG-Imediated antiviral responses to single-stranded RNA bearing 59-phosphates Replication-dependent potent IFN-alpha induction in human plasmacytoid dendritic cells by a single-stranded RNA virus 59-Triphosphate RNA is the ligand for RIG-I IRF-7 is the master regulator of type-I interferon-dependent immune responses Distinct and essential roles of transcription factors IRF-3 and IRF-7 in response to viruses for IFN-alpha/beta gene induction Interferon-beta is required for interferon-alpha production in mouse fibroblasts Viral infection switches non-plasmacytoid dendritic cells into high interferon producers Herpes simplex virus type-1 induces IFN-alpha production via Toll-like receptor 9-dependent and -independent pathways IFN regulatory factor 3-dependent induction of type I IFNs by intracellular bacteria is mediated by a TLR-and Nod2-independent mechanism Type I interferon production enhances susceptibility to Listeria monocytogenes infection Recognition of cytosolic DNA activates an IRF-3-dependent innate immune response A Toll-like receptor-independent antiviral response induced by double-stranded B-form DNA Distinct TLR-and NLR-mediated transcriptional responses to an intracellular pathogen DAI (DLM-1/ZBP1) is a cytosolic DNA sensor and an activator of innate immune response Regulation of innate immune responses by DAI (DLM-1/ZBP1) and other DNA-sensing molecules TANKbinding kinase-1 delineates innate and adaptive immune responses to DNA vaccines Recombinant adenoviral vectors turn on the type I interferon system without inhibition of transgene expression and viral replication Primary activation of interferon A and interferon B gene transcription by interferon regulatory factor 3 The effect of heat on the anatomy of the adenovirus Adenovirus efficiently transduces plasmacytoid dendritic cells resulting in TLR9-dependent maturation and IFN-alpha production Preferential activation of Toll-like receptor nine by CD46-utilizing adenoviruses Monitoring foreign gene expression by a human adenovirus-based vector using the firefly luciferase gene as a reporter Characterization of transgenic mice containing adenovirus early region 3 genomic DNA Recombinant adenovirus induces maturation of dendritic cells via an NF-kappaB-dependent pathway Mouse strain differences in plasmacytoid dendritic cell frequency and function revealed by a novel monoclonal antibody Antigen-driven induction of CD11c on intestinal intraepithelial lymphocytes and CD8+ T cells in vivo Regulated expression of the Mac-1, LFA-1, p150,95 glycoprotein family during leukocyte differentiation Development and function of murine B220+CD11c+NK1.1+ cells identify them as a subset of NK cells In vivo depletion of CD11c(+) dendritic cells abrogates priming of CD8(+) T cells by exogenous cell-associated antigens NK cell activation in visceral leishmaniasis requires TLR9, myeloid DCs, and IL-12, but is independent of plasmacytoid DCs Dendritic Cells Prime Natural Killer Cells by trans-Presenting Interleukin 15 TLR9-dependent recognition of MCMV by IPC and DC generates coordinated cytokine responses that activate antiviral NK cell function Herpes simplex virus type 1 activates murine natural interferon-producing cells through toll-like receptor 9 Toll-like receptor 9-mediated recognition of Herpes simplex virus-2 by plasmacytoid dendritic cells Innate antiviral responses by means of TLR7-mediated recognition of single-stranded RNA Recognition of double-stranded RNA and activation of NF-kappaB by Toll-like receptor 3 Functions of site-specific histone acetylation and deacetylation Sensing infection by adenovirus: Toll-like receptor-independent viral DNA recognition signals activation of the interferon regulatory factor 3 master regulator The RNA helicase RIG-I has an essential function in double-stranded RNAinduced innate antiviral responses Adenovirus-activated PKA and p38/MAPK pathways boost microtubulemediated nuclear targeting of virus Activation of p38 and ERK signaling during adenovirus vector cell entry lead to expression of the C-X-C chemokine IP-10 Interaction of adenovirus type 5 fiber with the coxsackievirus and adenovirus receptor activates inflammatory response in human respiratory cells Adenovirus infection stimulates the Raf/MAPK signaling pathway and induces interleukin-8 expression The adenovirus protease is required for virus entry into host cells Adenovirus serotype 7 retention in a late endosomal compartment prior to cytosol escape is modulated by fiber protein Subversion of CtBP1-controlled macropinocytosis by human adenovirus serotype 3 Adenovirus infection dramatically augments lipopolysaccharide-induced TNF production and sensitizes to lethal shock Increased expression of CD14 in macrophages after inhibition of the cholesterol biosynthetic pathway by lovastatin Toll-like receptor 4 expression levels determine the degree of LPS-susceptibility in mice Dendritic cell responses to early murine cytomegalovirus infection: subset functional specialization and differential regulation by interferon alpha/ beta Virus-induced interferon alpha production by a dendritic cell subset in the absence of feedback signaling in vivo Control of coronavirus infection through plasmacytoid dendritic-cell-derived type I interferon Modified vaccinia virus Ankara induces Toll-like receptor independent type I interferon responses Interferon alpha/beta and interleukin 12 responses to viral infections: pathways regulating dendritic cell cytokine expression in vivo Tissue-specific positive feedback requirements for production of type I interferon following virus infection Bone marrow stromal cell antigen 2 is a specific marker of type I IFN-producing cells in the naive mouse, but a promiscuous cell surface antigen following IFN stimulation Spatiotemporal regulation of MyD88-IRF-7 signalling for robust type-I interferon induction Viral DNA and a viral peptide can act as cofactors of adenovirus virion proteinase activity Double-stranded DNA and double-stranded RNA induce a common antiviral signaling pathway in human cells Essential role of IPS-1 in innate immune responses against RNA viruses Direct suppression of Stat1 function during adenoviral infection Contribution of interferon-beta to the murine macrophage response to the toll-like receptor 4 agonist, lipopolysaccharide Role of interferons in LPS hypersensitivity Myeloid and plasmacytoid dendritic cells are susceptible to recombinant adenovirus vectors and stimulate polyfunctional memory T cell responses Type I IFN innate immune response to adenovirus-mediated IFN-gamma gene transfer contributes to the regression of cutaneous lymphomas Targeted disruption of the MyD88 gene results in loss of IL-1-and IL-18-mediated function Identification of Lps2 as a key transducer of MyD88-independent TIR signalling A Toll-like receptor recognizes bacterial DNA The Unc93b1 mutation 3d disrupts exogenous antigen presentation and signaling via Toll-like receptors 3, 7 and 9 Differential contribution of Toll-like receptors 4 and 2 to the cytokine response to Salmonella enterica serovar Typhimurium and Staphylococcus aureus in mice Functional role of type I and type II interferons in antiviral defense An advanced culture method for generating large quantities of highly pure dendritic cells from mouse bone marrow Immunotherapy of patients with hormone-refractory prostate carcinoma pretreated with interferon-gamma and vaccinated with autologous PSA-peptide loaded dendritic cells-a pilot study Activation of the beta interferon promoter by unnatural Sendai virus infection requires RIG-I and is inhibited by viral C proteins Tumor necrosis factor alpha enhances influenza A virus-induced expression of antiviral cytokines by activating RIG-I gene expression Inquiring into the differential action of interferons (IFNs): an IFN-alpha2 mutant with enhanced affinity to IFNAR1 is functionally similar to IFN-beta Promotion of alpha/beta interferon induction during in vivo viral infection through alpha/ beta interferon receptor/STAT1 system-dependent and -independent pathways A type I interferon autocrine-paracrine loop is involved in Toll-like receptor-induced interleukin-12p70 secretion by dendritic cells Latency typespecific distribution of epigenetic marks at the alternative promoters Cp and Qp of Epstein-Barr virus The first step of adenovirus type 2 disassembly occurs at the cell surface, independently of endocytosis and escape to the cytosol We are grateful to C. Galanos, S. Tchapchet, P. Nielsen and I. Gyory