Research ArticleImmunology

Viral entry route determines how human plasmacytoid dendritic cells produce type I interferons

See allHide authors and affiliations

Science Signaling  03 Mar 2015:
Vol. 8, Issue 366, pp. ra25
DOI: 10.1126/scisignal.aaa1552

Abstract

Although plasmacytoid dendritic cells (pDCs) represent a rare immune cell type, they are the most important source of type I interferons (IFNs) upon viral infection. Phagocytosed RNA viruses and RNA virus–infected cells are detected by pDCs with the endosomal pattern recognition receptor (PRR) toll-like receptor 7 (TLR7). We showed that replication of the yellow fever live vaccine YF-17D in human pDCs and pDC-like cell lines stimulated type I IFN production through RIG-I (retinoic acid–inducible gene I), a member of the RIG-I–like receptor (RLR) family of cytosolic PRRs. Thus, human pDCs sense replicative viral RNA. In contrast, direct contact between pDCs and YF-17D–infected cells stimulated a TLR7-dependent, viral replication–independent production of type I IFN. We also showed that the RLR pathway was dampened by the activities of interleukin-1 receptor–associated kinases 1 and 4 (IRAK1 and IRAK4), which are downstream effectors of the TLR7 pathway, suggesting that both kinases play opposing roles downstream of specific PRRs. Together, these data suggest that a virus can stimulate either TLR or RLR signaling in the same cell, depending on how its nucleic acid content is delivered.

INTRODUCTION

The immune response to viral infection is initiated when pattern recognition receptors (PRRs) recognize specific nonself motifs in viral products, such as viral nucleic acids. These PRRs can be membrane-associated, such as toll-like receptors (TLRs), or cytosolic, such as retinoic acid–inducible gene I (RIG-I)–like receptors (RLRs) (1). Upon binding to viral nucleic acids, these receptors undergo conformational changes that enable their interaction with adaptor proteins and the recruitment of signaling complexes. These events lead to the activation of transcription factors, such as interferon regulatory factor 3 (IRF3) and IRF7, which stimulate the rapid expression of genes encoding the type I interferons (IFNs) IFN-α and IFN-β. Upon their secretion, IFN-α and IFN-β bind to their receptors (in an autocrine or paracrine manner) to activate the canonical Janus-activated kinase (JAK)–signal transducer and activator of transcription (STAT) signaling pathway. Subsequently, expression is induced in as many as 400 IFN-stimulated genes (ISGs), whose products have antiviral properties (2), effectively establishing the antiviral state in infected and surrounding cells. When exposed to viruses in vitro, most cells, even nonhematopoietic cells, produce type I IFNs.

Despite representing a rare cell type in the immune system, plasmacytoid dendritic cells (pDCs) are the most important source of type I IFN in response to viral infection (3, 4). The canonical model of pDC activation states that pDCs respond to RNA viruses exclusively through TLR7. This model is based on data generated from experiments with pDCs isolated from mice deficient in essential components of the RLR pathway, such as RIG-I, mitochondrial antiviral signaling protein (MAVS), tumor necrosis factor receptor–associated factor (TRAF)–associated nuclear factor κB (NF-κB) activator (TANK)–binding kinase 1 (TBK1), or inhibitor of κB (IκB) kinase ε (IKKε), which were then infected with RNA viruses (57). Many more studies of ex vivo murine pDCs or mouse models have corroborated these data (810), which were extrapolated to human pDCs.

Intrinsic viral replication is not required for TLR7-dependent recognition by pDCs. For example, in mouse pDCs, heat-inactivated (8), formaldehyde-fixed (11), or ultraviolet (UV)–damaged (12) influenza virus induces TLR7-dependent production of IFN-α similar to that with live virus, which suggests that the presence of viral RNA in TLR7-containing endosomal, lysosomal, or endolysosomal compartments is sufficient to activate a response. The accessibility of viral RNA to TLR7 might be a result of leaky viral fusion or the premature uncoating or degradation of incoming virions (8, 13, 14). Alternatively, cytosolic viral replication intermediates may be delivered to TLR7-containing lysosomes by the process of autophagy (12). TLR7-dependent stimulation of pDCs also occurs through direct physical contact between infected cells and pDCs (1521). This process involves the transfer of RNA-containing exosomes (15, 16) or viral particles (18, 19) between infected cells and pDCs.

The “TLR7-exclusive” sensing model has been challenged by experiments showing that recognition of human respiratory syncytial virus by human pDCs is replication-dependent but TLR7-independent (22, 23). Moreover, genetic studies have described human patients with TLR7 signaling deficiency who are surprisingly not predisposed to viral infection (24). Finally, human pDCs and pDC-like cell lines respond in an RIG-I–dependent manner to the cytosolic delivery of the poly-U/UC tract of the hepatitis C virus (HCV) genome (25) or synthetic double-stranded RNA (dsRNA) containing a 5′ triphosphate group (23). Together, these reports suggest that TLR-mediated recognition is probably not the exclusive mode of viral nucleic acid sensing in pDCs and that the RLR pathway might well be activated by RNA viruses in human pDCs.

Yellow fever virus (YFV) is an enveloped, positive-sense RNA virus belonging to the Flavivirus genus (26). In humans, YFV primarily targets the liver, but other tissues, such as the heart, kidneys, and lungs, are also sites of replication (27). Mouse models suggest that the vaccine strain of the virus has a similar tropism to that of its pathogenic counterpart (28). The live-attenuated YFV strain 17D-204 (YF-17D) was developed empirically in the 1930s by repeated passage of the virulent strain in animals and cultured cells. Vaccination of humans with YF-17D strongly induces the production of type I IFN as well as the expression of a network of ISGs (2931). Mechanisms underlying the sensing of YF-17D are unclear; the production of IFNs in response to many flaviviruses is mediated by both RLRs and TLRs in nonimmune cells (32). Here, we investigated the ability of YF-17D to stimulate human pDCs.

RESULTS

Infection of human pDCs by cell-free YF-17D stimulates the RLR-mediated production of type I IFN

We examined whether pDCs produce type I IFN when they encounter viral particles that are not associated with cells (so-called “cell-free” virus). Human peripheral blood mononuclear cells (PBMCs) and pDC-depleted PBMCs were infected with infectious YF-17D or with nonreplicative, UV-inactivated YF-17D for 24 hours. The abundances of IFNA and IFNB mRNAs were increased in infected PBMCs compared to those in uninfected cells but were unchanged in pDC-depleted PBMCs (Fig. 1, A and B, and fig. S1), which suggests that pDCs represent the main source of type I IFNs in human PBMCs after exposure to YF-17D. Consistently, IFN-α protein was secreted by infected PBMCs but not by infected PBMCs depleted of pDCs (Fig. 1C). When exposed to UV-treated virions, neither group of cells exhibited substantial increases in type I IFNs at either the mRNA or protein level (Fig. 1, A to C). Intact PBMCs produced about 10-fold more viral RNA than did PBMCs depleted of pDCs (Fig. 1D), which suggests that YF-17D might replicate in pDCs. Viral RNA detected in cells exposed to UV-inactivated virions represented the amount of nonreplicative viral RNA that penetrated the cells (Fig. 1D).

Fig. 1 Infection of human pDCs by cell-free YF-17D stimulates the RLR-dependent production of type I IFN.

(A to D) Freshly isolated PBMCs or pDC-depleted PBMCs were left uninfected (mock) or were infected with live or UV-inactivated (UV-inact.) YF-17D for 24 hours. The relative amounts of IFNA mRNA (A) and IFNB mRNA (B) were determined by quantitative polymerase chain reaction (qPCR) analysis. (C) Cell culture medium was analyzed by enzyme-linked immunosorbent assay (ELISA) to determine the amounts of IFN-α secreted by the indicated cells. #, values are below the limit of detection of the IFN-α ELISA assay (12.5 pg/ml). (D) The relative amounts of cell-associated viral RNA were determined by qPCR analysis and were expressed as genome equivalents (GE) per microgram of total cellular RNA. All results are expressed as fold increases relative to uninfected cells. (E to G) Freshly isolated pDCs were left untreated or were treated with MRT67307, IRS661, or IRS1040 for 30 min. The cells were then infected with UV-inactivated or live YF-17D for 24 hours, as indicated. (E) The relative amounts of cell-associated viral RNA were determined by qPCR analysis (black bars). The amounts of viral RNA in cell culture media were also determined by qPCR analysis and were expressed as GE per milliliter of culture medium (gray bars). (F) The relative amounts of IFNA and IFNB mRNAs were determined by qPCR analysis. (G) Cell culture medium was analyzed by ELISA to determine the amounts of IFN-α secreted by the indicated cells. Data in (A) to (G) are means ± SD of at least three independent experiments.

Infection of purified human pDCs with YF-17D resulted in the intracellular production of viral RNA (Fig. 1E and fig. S1). Moreover, the culture medium of pDCs infected with live YF-17D contained ~1000-fold more viral RNA than did the culture medium of pDCs exposed to UV-inactivated YF-17D (Fig. 1E). Detection of viral RNA in the culture medium of cells exposed to UV-inactivated virions represented virus particles that did not bind to cell membranes. These data suggest that YF-17D replicated in human pDCs and produced virions. Cells treated before infection with MRT67307, an inhibitor of the RLR-associated kinases TBK1 and IKKε (3335), produced substantially more viral RNA than did cells treated with vehicle before infection (Fig. 1E). In contrast, viral replication was not affected by pretreatment of the pDCs with IRS661, a specific inhibitor of TLR7 (36, 37), or IRS1040, an oligodeoxynucleotide that acts as a negative control for IRS661 (Fig. 1E). The amount of viral RNA in the culture medium of MRT67307-treated cells was increased, albeit not statistically significant, compared to that in the medium of vehicle-treated cells (Fig. 1E).

The amounts of type I IFN mRNA and protein generated by infected cells pretreated with MRT67307 were reduced compared to those generated by vehicle-treated cells (Fig. 1, F and G). Neither IRS661 nor IRS1040 had any effect on the generation of type I IFN mRNA or protein (Fig. 1, F and G). When pDCs were treated with the TLR7 agonist CL264, the generation of type I IFN mRNA and protein was impaired by IRS661 (fig. S2, A and B), demonstrating that IRS661 was effective in human purified pDCs. Purified human pDCs produce type I IFNs when they encounter YF-17D (38). Here, we showed that the virus replicates in pDCs and induces the production of type I IFN in a TBK1- and IKKε-dependent but TLR7-independent manner. Although this response was limited (compare Fig. 1G with fig. S2B), it was sufficient to substantially affect viral replication (Fig. 1E).

RIG-I directly senses replicative YF-17D in pDC-like cells

To identify the cytosolic sensor involved in the recognition of YF-17D RNA in pDCs, we performed gene-silencing experiments with Gen2.2 cells, which act as a reliable surrogate for pDCs (18, 25, 39). Viral RNA was detected in the culture medium of Gen2.2 cells that were infected for 24 hours with YF-17D (Fig. 2A). The use of a large number of cells enabled us to titrate the virus by plaque assays, which showed that the virions released by the Gen2.2 cells were infectious (Fig. 2B). As expected, Gen2.2 cells exposed to UV-treated virions did not release infectious virus particles (Fig. 2B). Through immunofluorescence analysis, we found that ~20% of Gen2.2 cells stained positive either for the nonstructural viral protein NS1, which is not carried in virions, or for dsRNA, a product synthesized during viral replication (Fig. 2C). Flow cytometric analysis showed that, on average, 12% of cells were positive for dsRNA (Fig. 2D). Therefore, depending on the sensitivity of the assay, 12 to 20% of Gen2.2 cells were infected by YF-17D. By comparison, ~70% of Vero cells (a monkey kidney epithelial cell line) contained dsRNA 24 hours after infection (fig. S3, A and B). The low permissivity of Gen2.2 cells for YF-17D could be explained by these cells having decreased amounts of factors critical for viral internalization, replication, or both or an increased abundance of pDC-specific viral restriction factors. YF-17D replication stimulated a statistically significant increase in the amounts of type I IFN mRNAs in infected Gen2.2 cells compared to those in uninfected cells (Fig. 2E), as well as in the secretion of IFN-α (Fig. 2F). Therefore, similar to pDCs, Gen2.2 cells are productive targets of YF-17D and produce type I IFN in response to infection. Note that the amount of IFN-α secreted by infected Gen2.2 cells was similar to that secreted by YF-17D–infected pDCs, despite the larger number of cells that were infected (compare Fig. 2F and Fig. 1C). This is consistent with previous reports suggesting that Gen2.2 cells are less potent producers of IFNs than are pDCs (18, 19).

Fig. 2 RIG-I directly senses replicating YF-17D.

(A to F) Gen2.2 cells were left uninfected (mock) or were infected with live or UV-inactivated YF-17D for 24 hours. (A) The relative amounts of cell-associated viral RNA as well as viral RNA in cell culture media were determined by qPCR analysis. (B) The presence of extracellular infectious viruses in culture medium of the indicated cells was determined by plaque assays with Vero cells. PFU, plaque forming units. (C) Infected Gen2.2 cells were stained with anti-NS1 or anti-dsRNA antibodies (green) and phalloidin (red) to detect F-actin and 4′,6-diamidino-2-phenylindole (DAPI; blue) to detect nuclei. Images are representative of 200 observations from three independent experiments. (D) The percentages of the indicated cells that contained dsRNA were determined by flow cytometric analysis. (E) The relative amounts of IFNA and IFNB mRNAs were determined by qPCR analysis and were expressed as the fold increase in abundance relative to those in uninfected cells. (F) Cell culture media were analyzed by ELISA to determine the amounts of IFN-α secreted by indicated cells. (G) Top: Gen2.2 cell lines established by transduction with lentiviruses expressing Scrambled (Scr.) shRNA or RIG-I– or TLR7-specific shRNAs were subjected to Western blotting analysis with antibodies against indicated proteins. Bottom: Quantitative analysis of the extent of shRNA-mediated knockdown of RIG-I and TLR7. (H to J) The indicated shRNA-expressing Gen2.2 cell lines were infected with YF-17D for 24 hours. (H) Cell-associated viral RNA was measured by qPCR analysis. (I) The relative abundances of IFNA and IFNB mRNAs were determined by qPCR analysis. Results are presented as percentages of the amounts of the indicated mRNAs detected in infected Gen2.2-shScrambled cells, which were set at 100%. (J) Cell culture media were analyzed by ELISA to determine the amounts of IFN-α secreted by the indicated cells. (K) Gen2.2-shScrambled and Gen2.2-shRIG-I cells were left uninfected (mock) or were infected with YF-17D at a multiplicity of infection of 50 for 24 hours. Whole-cell lysates were then analyzed by Western blotting with antibodies against the indicated proteins. Western blots are representative of two independent experiments. For densitometric analysis of band intensities, see fig. S2, D to G. Data in all bar graphs are means ± SD of three independent experiments.

We next assessed the presence of RIG-I, MDA5, and TLR7 mRNAs in Gen2.2 cells and pDCs. Hepatocarcinoma Huh7 cells, which are thought to express all three sensors (40, 41), were used as control cells. Both pDCs and Gen2.2 cells exhibited similar expression profiles for all three sensors (fig. S3C). Contrary to a previous report (40), we did not detect TLR7 mRNA in Huh7 cells. Transduction of Gen2.2 cells with lentiviruses expressing short hairpin RNAs (shRNAs) specific for either RIG-I or TLR7 to generate the Gen2.2-shRIG-I and Gen2.2-shTLR7 cell lines, respectively, reduced the abundances of both proteins by ~70% compared to those in Gen2.2 cells transduced with lentivirus expressing scrambled control shRNA (Gen2.2-shScrambled cells) (Fig. 2G). Viral replication in Gen2.2-shTLR7 was not markedly different from that in Gen2.2-shScrambled cells 24 hours after infection (Fig. 2H). By contrast, Gen2.2-shRIG-I cells produced substantially more viral RNA than did Gen2.2-shScrambled cells (Fig. 2H). This is consistent with our earlier data showing that inhibition of the RLR-associated kinases TBK1 and IKKε in YF-17D–infected pDCs substantially increased the extent of viral replication (Fig. 1E). The Gen2.2-shRIG-I cells generated ~70% less IFNA mRNA and ~60% less IFNB mRNA compared with Gen2.2-shScrambled control cells when infected by cell-free virions (Fig. 2I). Consistent with these data, IFN-α was secreted by Gen2.2-shScrambled and Gene2.2-shTLR7 cells, but not by Gen2.2-shRIG-I cells, upon infection by cell-free virions (Fig. 2J). Note that none of the shRNA-expressing cell lines secreted IFN-α if they were left uninfected (Fig. 2J). Moreover, upon infection, the RIG-I downstream kinase TBK1 and the transcription factor IRF3, which are typically activated by RLR signaling, were both ~60% less phosphorylated in Gen2.2-shRIG-I cells than in control cells, despite their total protein abundances in both cell lines being similar (Fig. 2K and fig. S3, D and E). The transcription factor STAT1 was activated in Gen2.2-shScrambled cells, but not Gen2.2-shRIG-I cells, upon infection by cell-free virions (Fig. 2K and fig. S3F), demonstrating the effectiveness of the RIG-I pathway in Gen2.2 cells. As expected for an ISG, RIG-I abundance was increased by cell-free virions in both Gen2.2-shScrambled control cells and Gen2.2-shRIG-I cells (Fig. 2K and fig. S3G).

Immunofluorescence analysis showed that both IRF3 and IRF7 were localized to the cytoplasm in 100% of uninfected cells (fig. S4). IRF3 was found in both the cytosol and the nucleus of 71% of infected cells, whereas IRF7 was detected exclusively in the cytosol of 97% of infected cells (fig. S4). These data suggest that the RIG-I–mediated response to viral infection led to the activation of IRF3 only. Thus, the replication of YF-17D activates the RIG-I–TBK-1–IRF3 axis in Gen2.2 cells, supporting the data obtained from our earlier experiments with pDCs (Fig. 1, F and G). We only assessed the requirement for RIG-I in Gen2.2 cells infected with YF-17D; however, according to results obtained from a study of human embryonic kidney (HEK) 293 cells (29), it is likely that other RLRs [for example, Mda5 (melanoma differentiation–associated protein 5)] are involved in the sensing of YF-17D.

Sendai virus–stimulated production of IFN in Gen2.2 cells is RIG-I–dependent

To extend our observation that the RLR pathway is activated by viral RNA in Gen2.2 cells, we infected these cells with Sendai virus (SeV), a negative-sense RNA virus of the family Paramyxoviridae. SeV activates RIG-I in nonhematopoietic cells (42, 43) and is a potent inducer of type I IFN production (44). Eight hours after infection with SeV, Gen2.2-shRIG-I cells generated decreased amounts of IFNA and IFNB mRNAs compared to those generated by SeV-infected control cells (Fig. 3, A and B). On the other hand, the amounts of IFNA and IFNB mRNAs in Gen2.2-shRIG-I cells treated with the TLR7 agonist CL264 were similar to those produced by CL264-treated control cells (Fig. 3, A and B). As expected, cells infected with SeV or treated with CL264 exhibited increased amounts of RIG-I protein compared to those in uninfected and untreated control cells (Fig. 3C and fig. S5A). Furthermore, both TBK1 and IRF3 were activated in SeV-infected Gen2.2-shScrambled control cells but not in Gen2.2-shRIG-I cells (Fig. 3C and fig. S5, B and C). These results suggest that SeV-mediated induction of type I IFN in Gen2.2 cells is RIG-I–dependent. Note that the TLR7 agonist CL264 failed to stimulate IRF3 phosphorylation (Fig. 3C), which suggests that this transcription factor is not activated downstream of TLR7. The abundance of STAT1 protein was increased in cells exposed to SeV or treated with CL264 compared to that in untreated control cells (Fig. 3C and fig. S5D). This increase in STAT1 abundance was abolished in SeV-infected Gen2.2-shRIG-I cells (Fig. 3C and fig. S5D). Similarly, there was a ~50% decrease in STAT1 phosphorylation in SeV-infected Gen2.2-shRIG-I cells compared to that in control cells, but there was no effect of loss of RIG-I on CL264-dependent STAT1 phosphorylation (Fig. 3C and fig. S5E). Together, these results suggest that the SeV-mediated induction of type I IFN in Gen2.2 cells is RIG-I–dependent. Therefore, RLR signaling in pDC-like cells is not restricted to a particular family of viruses.

Fig. 3 Sendai virus stimulates RIG-I signaling in Gen2.2 cells.

(A to C) Gen2.2-shScrambled and Gen2.2-shRIG-I cells were left uninfected (mock), were infected with SeV, or were treated with CL264 for 8 hours. The relative amounts of IFNA mRNA (A) and IFNB mRNA (B) were determined by qPCR analysis and normalized to that of GAPDH mRNA. Results are expressed as percentages of the abundances of IFNA mRNA and IFNB mRNA in the indicated Gen2.2-shScrambled cells. Data are means ± SD of three independent experiments. (C) Whole-cell lysates of the indicated cells were analyzed by Western blotting with antibodies against the indicated proteins. Western blots are representative of two independent experiments. For densitometric analysis of band intensities, see fig. S5.

pDCs are stimulated by direct contact with YF-17D–infected cells

pDCs can be stimulated by physical contact with cells infected by various viruses (1520). We therefore investigated whether pDCs recognized YF-17D–infected cells. We used Vero cells as producer cells because they lack functional IFNA- and IFNB-encoding genes (45), which ensures that pDCs are the only source of type I IFNs in mixed cultures of Vero cells and pDCs. Vero cells, which are adherent, were infected for 16 hours and then were cocultured for 24 hours with human PBMCs or pDC-depleted PBMCs. The amount of viral RNA produced by Vero cells cocultured with PBMCs was similar to that of Vero cells cocultured with pDC-depleted PBMCs (Fig. 4A), which suggests that Vero cells were the main producers of viral RNA in this context. IFNA and IFNB mRNAs were abundant in Vero cells cocultured with PBMCs but not in Vero cells cocultured with pDC-depleted PBMCs (Fig. 4, B and C). Consistently, ~300 pg/ml of IFN-α was produced by cocultured Vero cells and PBMCs, but no IFN-α was produced by cocultures of Vero cells with pDC-depleted PBMCs (Fig. 4D) These data suggest that pDCs represent the only source of type I IFNs among human PBMCs stimulated by YF-17D–infected cells.

Fig. 4 YF-17D–infected cells induce the production of type I IFN by human pDCs.

(A to D) Vero cells were left uninfected (mock) or were infected with live YF-17D virus or UV-inactivated YF-17D virus for 16 hours. The Vero cells were then cocultured with human PBMCs or pDC-depleted human PBMCs for a further 24 hours. The relative amounts of cell-associated viral RNA (A), IFNA mRNA (B), and IFNB mRNA (C) were determined by qPCR analysis as described earlier. (D) Cell culture media were analyzed by ELISA to determine the amount of IFN-α secreted by the indicated cells. (E to H) Vero cells were left uninfected (mock) or were infected with live YF-17D virus or UV-inactivated YF-17D for 16 hours. The Vero cells were then cocultured with freshly isolated pDCs for a further 24 hours (gray bars). In parallel, pDCs (in the absence of Vero cells) were left uninfected (mock) or were infected with live or UV-inactivated YF-17D virus for 24 hours. Relative amounts of cell-associated viral RNA (E), IFNA mRNA (F), and IFNB mRNA (G) were determined by qPCR analysis as described earlier. (H) Cell culture media were analyzed by ELISA to determine the amount of IFN-α secreted by the indicated cells. Data in all bar graphs are means ± SD or three independent experiments.

We then performed experiments in which purified human pDCs were cocultured with Vero cells. As expected, the amount of viral RNA was greater in Vero cells cocultured with pDCs than in pDCs cultured alone (Fig. 4E). Furthermore, pDCs cocultured with Vero cells contained greater amounts of IFNA and IFNB mRNAs than pDCs cultured alone (Fig. 4, F and G). Cocultures of Vero cells and pDCs contained about 10 times more secreted IFN-α compared with cultures of pDCs alone (Fig. 4H). Together, these data suggest that YF-17D–infected cells are potent stimulators of pDCs (Fig. 4), whereas cell-free virions are modest inducers of type I IFN production by pDCs (Fig. 1).

Gen2.2 cells respond to YF-17D–infected cells through direct cell-to-cell contact

We then used Gen2.2 cells to further investigate the sensing of YF-17D–infected cells. Vero cells were infected with YF-17D for 16 hours and then were cocultured for 24 hours with carboxyfluorescein succinimidyl ester (CFSE)–labeled Gen2.2 cells, which, similar to pDCs, grow in suspension. After gentle shaking of the mixed culture, we analyzed both adherent cells (mostly Vero cells) and suspension cells (mostly Gen2.2 cells). Microscopic observations revealed that the adherent cells were mostly composed of Vero cells, some of which were in close contact with Gen2.2 cells (Fig. 5A). We found that 75% of the Vero cells contained viral products (Fig. 5A), which was confirmed by flow cytometric analysis (Fig. 5, B and C). On the other hand, the CFSE-labeled Gen2.2 cells that were in contact with infected Vero cells had undetectable NS1 or viral dsRNA (Fig. 5A). The low number (around 5%) of Gen2.2 cells that were infected in the coculture are likely to have been suspension cells that were infected by the newly released YF-17D virus (Fig. 5, B and C).

Fig. 5 Gen2.2 cells respond to YF-17D-infected cells by establishing a direct cell-to-cell contact.

(A to H) Vero cells were left uninfected (mock) or were infected for 16 hours with YF-17D virus and then were cocultured with Gen2.2 cells for a further 24 hours. As indicated, either Vero cells or Gen2.2 cells were labeled with CFSE before infection or coculture. (A) CFSE-labeled Gen2.2 cells (green) and Vero cells were stained with antibodies specific for viral NS1 protein (red, top) or dsRNA (red, bottom) and DAPI to visualize nuclei (blue). Images are representative of three independent experiments. (B) CFSE-labeled Vero cells and Gen2.2 cells were stained with antibody specific for YF-17D envelope protein and then were analyzed by flow cytometry. Data are representative of three independent experiments. (C) Analysis of the percentages of cells expressing the YF-17D envelope protein from cocultures (gray bars) and individual cultures of Gen2.2 and Vero cells that were infected with YF-17D for 24 hours (black bars). Data are means ± SD of three independent experiments. (D to G) Vero cells and Gen2.2 cells were cocultured in wells in which the two cell types were or were not separated by transwell chambers, as indicated. Suspension cells (predominantly Gen2.2 cells) were collected after gentle shaking of the wells, whereas adherent cells (predominantly Vero cells) were collected separately. Both sets of cells were then subjected to qPCR analysis to determine relative amounts of cell-associated viral RNA (D), IFNA mRNA (E), and IFNB mRNA (F). (G) Cell culture media were analyzed by ELISA to determine the amounts of IFN-α secreted by cells under the indicated conditions. (H) CLEM analysis of the contact between YF-17D–infected Vero cells and Gen2.2 cells. Cells were fixed and stained for viral envelope protein (green), F-actin (red), and DAPI (blue). Cells were imaged with an epifluorescence microscope and were then processed for electron microscopic analysis. Enlarged views of the top left image are shown in the lower set of panels and are representative of two independent experiments. Note that the bottom left panel shows a slightly titled view of the central panel. (I and J) YF-17D–infected Vero cells, YFRP cells, and HCVRP cells were cultured alone or in coculture with Gen2.2 cells for 24 hours. Cells were then subjected to qPCR analysis to determine relative amounts of cell-associated viral RNAs (I) or IFNA and IFNB mRNAs (J). Data are means ± SD of four independent experiments.

We detected similar amounts of viral RNA in the adherent cells and the suspension cells (Fig. 5D). Because Gen2.2 cells were poorly infected with YF-17D virus (Fig. 5, B and C), it is likely that the main producers of viral RNA among the suspended cells were Vero cells that had become detached upon shaking. Consistent with this, Gen2.2 cells that were cultured in the same wells as Vero cells but were separated by a transwell chamber generated substantially decreased amounts of viral RNA (Fig. 5D). When cell-to-cell contacts were permitted (in the absence of transwells), adherent cells were potent inducers of IFNA, whereas suspension cells were modest inducers (Fig. 5E). Adherent cells generated more IFNB mRNA than did suspended cells when cell-to-cell contacts were permitted (Fig. 5F), although the amounts generated were lower than those of IFNA mRNA. About 1 ng/ml of IFN-α was released by Gen2.2 cells stimulated by infected Vero cells, but only when there was contact between the cells (Fig. 5G). Thus, Gen2.2 cells that were attached to Vero cells were robust inducers of type I IFNs.

To ensure that the production of type I IFN by Gen2.2 cells in response to contact with YF-17D–infected cells was not cell type–specific, we cocultured Gen2.2 cells with hepatocarcinoma Huh7 cells, which are adherent cells (fig. S6, A to D). Moreover, in humans, YFV primarily targets the liver and replicates in hepatocytes (27). Upon infection by cell-free YF-17D, the amounts of viral RNA and IFNA mRNA produced by cultures of Huh7 cells were greater than those of cultured Gen2.2 cells (fig. S6, A and B). When cell-to-cell contact occurred, the amounts of IFNA mRNA generated by both adherent and suspension cells substantially increased (fig. S6D). When Huh7 cells were cocultured with Gen2.2 cells but were separated by a transwell, Huh7 cells generated similar amounts of IFNA mRNA as they did when they were cultured alone (fig. S6, B and D). When Gen2.2 cells were cocultured with Huh7 cells, separated or not by a transwell, they generated greater amounts of IFNA mRNA than they did when they were cultured alone (fig. S6, B and D). This is most likely a result of secondary IFN responses, whereas the attachment of Gen2.2 cells to Huh7 cells was likely responsible for the enhanced generation of IFNA mRNA in adherent cells. These data suggest that Gen2.2 cells attached to Huh7 cells or Vero cells are robust inducers of type I IFN responses (Fig. 5, D to G, and fig. S6). Together, these results suggest that, similar to primary human pDCs, Gen2.2 cells respond to YF-17D–infected cells through cell-to-cell contact. Such responses did not appear to require viral replication in Gen2.2 cells (Fig. 5, A and B).

To further characterize the nature of the physical contact between YF-17D–infected Vero cells and Gen2.2 cells, we performed correlative light and electron microscopic (CLEM) analyses. Light microscopic analysis ensured that the Vero cells were infected (Fig. 5H). When viewed with an electron microscope, both cell types were easily identifiable by shape and size (Fig. 5H). Multiple observations revealed that infected Vero cells and Gen2.2 cells were connected by an intricate network of structures, including filopodia-like filaments (Fig. 5H). Because tight contact with infected cells is required for the stimulation of pDCs (Fig. 5, D to G), these connections might provide the conduit through which viral material is transferred.

The transfer of viral RNA from infected cells to pDCs has been reported for several RNA viruses, including HCV (15, 17). Such RNA transfer does not require the production of viral particles and might occur through exosomes (15). To determine whether the generation of YF-17D RNA in infected Vero cells was sufficient to stimulate the production of type I IFNs in Gen2.2 cells, we generated a Vero cell line stably expressing YF-17D replicons (YFRP cells). This cell line does not produce viral particles because only nonstructural proteins are expressed by the replicon. Huh7 cells stably expressing an HCV replicon (HCVRP cells) (46) were used as positive controls. YFRP cells produced on average 1 × 109 viral GE per microgram of total RNA (Fig. 5I). The YFRP cells were unable to stimulate the expression of IFNA or IFNB in cocultured Gen2.2 cells (Fig. 5I). HCVRP cells did not exhibit induction of IFNA or IFNB expression when cultured alone (Fig. 5J). By contrast, and as expected (15, 17), HCVRP cells induced the expression of IFNA and IFNB in cocultured Gen2.2 cells (Fig. 5J), despite exhibiting low levels of viral replication (Fig. 5I). Together, these data suggest that Gen2.2 cells, similar to pDCs (15, 17), are responsive to HCV RNA–producing cells. Because YFRP cells produced 1000-fold more viral RNA than did HCVRP cells (Fig. 5I) but did not stimulate Gen2.2 cells (Fig. 5J), it is likely that it is necessary for the infected cells to produce infectious viral particles to stimulate pDCs.

The sensing of YF-17D–infected cells by pDCs and Gen2.2 cells is TLR7-dependent

To identify the PRR in pDCs involved in the sensing of YF-17D–infected cells, we cultured purified human pDCs with YF-17D–infected Vero cells in the presence or absence of MRT67307 [an inhibitor of both TBK1 and IKKε (3335)], the specific TLR7 inhibitor IRS661 (36, 37), or the nonspecific control IRS1040. Of these, only IRS661 caused a substantial increase in the production of YF-17D RNA by the infected cells compared to that by vehicle-treated infected cells (Fig. 6A). Consistently, IRS661 resulted in a substantial reduction in the expression of IFNA and IFNB (Fig. 6B) and in the secretion of IFN-α (Fig. 6C), which suggests that YF-17D–infected cells stimulated pDCs to produce type I IFN in a TLR7-dependent manner. Knockdown of TLR7 in Gen2.2 cells slightly, but not statistically significantly, increased the extent of viral replication in cocultures with Vero cells after 24 hours (Fig. 6D). When infected Vero cells were cocultured with Gen2.2-shTLR7 cells, the abundances of IFNA and IFNB mRNAs were markedly reduced compared to those in cocultures with Gen2.2-shScrambled cells or Gen2.2-shRIG-I cells (Fig. 6E). Consistent with these data, IFN-α production was completely abolished in the coculture of YF-17D–infected Vero cells with Gen2.2-shTLR7 cells (Fig. 6F). Together, these results suggest that the stimulation of pDCs and Gen2.2 cells by YF-17D–infected cells depends on TLR7 in pDCs and Gen2.2 cells.

Fig. 6 The sensing of YF-17D–infected cells by pDCs and Gen2.2 cells is TLR7-dependent.

(A to C) Vero cells were infected with YF-17D for 16 hours and then were cocultured with freshly purified pDCs for a further 30 hours in the absence or presence of MRT67307, IRS661, or IRS1040. (A) The relative amounts of cell-associated viral RNA were determined by qPCR analysis. (B) The relative amounts of IFNA and IFNB mRNAs were determined by qPCR analysis and were expressed as normalized percentages relative to those of cells that were not treated with the inhibitor. (C) Cell culture media was analyzed by ELISA to determine the amounts of IFN-α secreted by uninfected or infected cells treated with the indicated inhibitors. (D to F) The indicated shRNA-expressing Gen2.2 cell lines were cocultured with YF-17D–infected Vero cells for 24 hours. The cells were then subjected to qPCR analysis to determine relative amounts of cell-associated viral RNA (D) and IFNA and IFNB mRNAs (E). (E) Results are expressed as percentages of the amounts of IFNA and IFNB mRNAs of infected Gen2.2-shScrambled cells, which were set at 100%. (F) Cell culture media was analyzed by ELISA to determine the amounts of secreted IFN-α. Data in all bar graphs are means ± SD of three independent experiments.

We next monitored the nuclear translocation of IRF3 and IRF7 in Gen2.2 cells cocultured with CFSE-labeled Vero cells (fig. S7) or, by comparison, in CFSE-labeled Vero cells alone (fig. S8). In uninfected Vero cells, cultured alone or with Gen2.2 cells, IRF3 exhibited faint, diffuse cytosolic staining (figs. S7A and S8A), whereas IRF7 was not detectable (figs. S7B and S8B). This finding was expected, given that IRF7 is not constitutively expressed in cells other than pDCs. IRF3 and IRF7 were both found in the cytosol of Gen2.2 cells that were in contact with uninfected Vero cells (fig. S7, A and B). IRF3 was present in the nucleus of 95% of infected Vero cells, whereas IRF7 remained undetectable (fig. S8). This suggests that in the context of direct infection by cell-free viruses, only IRF3 is activated in Vero cells. IRF3 remained cytosolic in 90% of Gen2.2 cells that were in close contact with Vero cells that exhibited nuclear localization of IRF3 (fig. S7A). Upon infection, IRF7 was found in the nucleus in 100% of Vero cells as well as in 83% of Gen2.2 cells that were in close contact with Vero cells (fig. S7B). Because IRF7 was not detectable in infected Vero cells cultured alone (fig. S8B), its expression and nuclear translocation in infected Vero cells in coculture suggested that it may have been induced by some factors, such as type I IFN released by Gen2.2 cells. Together, these results suggest that the large amount of type I IFN secreted by Gen2.2 cells that were in close contact with infected Vero cells depended on cell-to-cell contact, TLR7, and IRF7.

The kinase activities of IRAK1 and IRAK4 inhibit the RLR signaling pathway in pDCs

Numerous mechanisms mediate the inhibition of RLR signaling (47). One of these involves interleukin-1 receptor–associated kinase 1 (IRAK1), a kinase downstream of TLR7 whose inactivation increases RIG-I–mediated type I IFN responses in nonhematopoietic cells (48). IRAK1 is phosphorylated in resting cells and, in response to stimulation, is further phosphorylated, at least in monocytic cell lines (49, 50). To investigate the role of IRAK1 in RLR signaling in Gen2.2 cells and pDCs upon infection with YF-17D, we took advantage of a widely used inhibitor of the kinase activities of IRAK1 (35). This inhibitor also affects the kinase activity of IRAK4, which is the second active kinase of the IRAK complex in the TLR7 signaling pathway (51). Viral replication in Gen2.2 cells alone or in coculture with Vero cells was similar in the presence of the IRAK1/4 inhibitor to that in dimethyl sulfoxide (DMSO)–treated cells 24 hours after infection (Fig. 7A). Similar results were obtained from experiments with purified pDCs (Fig. 7A). Consistent with our TLR7-silencing experiments (Fig. 6, D to F), treatment with the IRAK1/4 inhibitor substantially decreased induction by YF-17D–infected Vero cells of IFNA and IFNB expression in Gen2.2 and pDCs and their secretion of IFN-α (Fig. 7, B to D). These results further suggest that the sensing of YF-17D–infected cells by pDCs activates the TLR7 pathway, consistent with our earlier results (Fig. 6). The IRAK1/4 inhibitor had a positive effect on the responses of Gen2.2 cells and pDCs after infection with cell-free virus (Fig. 7, B to D), which suggests that the kinase activity of IRAK1, IRAK4, or both inhibits the RLR signaling pathway in pDCs.

Fig. 7 The kinase activities of IRAK1 and IRAK4 inhibit the RLR signaling pathway in pDCs.

(A to D) Gen2.2 cells or freshly purified pDCs were infected with YF-17D or were cocultured with YF-17D–infected Vero cells for 24 hours in the presence of DMSO or an IRAK1/4 inhibitor (inh.). The indicated cells were then subjected to qPCR analysis to determine the relative amounts of cell-associated viral RNA (A), IFNA mRNA (B), and IFNB mRNA (C). Data are expressed as percentages of the indicated mRNAs in DMSO-treated cells. (D) Culture media of the indicated cells were analyzed by ELISA to determine the amounts of secreted IFN-α. Data in all bar graphs are means ± SD of three independent experiments.

DISCUSSION

By being able to rapidly produce up to 1000-fold more type I IFN than any other cell type, pDCs are unique players in the immune system (3, 4). Many experiments performed with ex vivo mouse pDCs deficient in components of the RLR pathway or with mouse models have led to the conclusion that pDC-mediated responses to RNA viruses are mediated exclusively through TLR7 (3, 4). The question that arises from this model is whether human pDCs are equipped to sense RNA viruses that replicate in their cytoplasm. Here, we showed that replication of a live virus stimulated RIG-I–mediated responses in pDCs. The Gen2.2 cell line enabled us to directly demonstrate the requirement for RIG-I in the sensing of YF-17D, which was confirmed in experiments with primary pDCs treated with an inhibitor of kinases downstream of RLRs. Infection of Gen2.2 cells with SeV, which is unrelated to YFV, also stimulated the RIG-I–dependent production of type I IFN, which suggests that RIG-I–mediated sensing is not restricted to a single family of viruses. Infection by either YF-17D or SeV activated TBK1 and IRF3, which suggests that the entire RIG-I–IRF3 axis is activated in pDCs. The activation of IRF3 by viral RNA appeared to be restricted to the RIG-I pathway because the TLR7 agonist CL264 did not stimulate the phosphorylation of IRF3. Consistent with this finding, IRF3, but not IRF7, was localized to the nucleus upon infection of Gen2.2 cells with YF-17D. Thus, these data suggest that RIG-I is capable of inducing type I IFN production upon viral replication in human pDCs. The presence of a functional RLR pathway in pDCs upon viral replication might have been missed in previous studies given their low permissivity to most viruses.

Our results also suggest that stimulation of pDCs and Gen2.2 cells by YF-17D–infected cells is TLR7-dependent. These results were expected on the basis of previous reports of experiments with various viruses (1520). However, in contrast to HCV and the classical swine fever virus (15, 17, 20, 21), which belong to the Flaviviridae family of which YFV is a member, expression of YF-17D RNA in target cells was not sufficient to stimulate type I IFN production in Gen2.2 cells. Thus, the mechanism exploited by YF-17D to spread from infected cells to pDCs may require the production of viral particles by infected cells and therefore may resemble the mechanism described for retroviruses (18, 19). The mechanism of cell-to-cell transfer of virions is not well characterized. Future studies will investigate whether newly assembled viruses are transferred from infected cells to pDCs through the physical connections observed by CLEM. These viruses might eventually fuse with TLR7-positive lysosomal compartments in which degradation of virions would render the viral RNA accessible to TLR7. In contrast, the RNA of penetrating, cell-free YF-17D virions is protected by its capsid during its transit through endosomal compartments (52), and it is therefore inaccessible to TLR7 (fig. S9). Our work is in agreement with data describing a negative effect of IRS661 on IFN-α production by human PBMCs infected with YF-17D (53). This mixed cell population consisted of several cell types, only 0.4% of which were pDCs. Therefore, in this previous study (53), IFN-α may have been produced mainly by pDCs that sensed other infected blood cells through TLR7.

The disparity between the intensities of the TLR7- and RIG-I–mediated IFN responses in pDCs has been observed upon ligand stimulation and was suggested to be a result of the lower basal abundance of RIG-I compared to that of TLR7 (54). However, our results would suggest that a dependency on viral replication, which occurs to a low extent in pDCs, may be responsible for the decreased magnitude of the RIG-I–dependent response compared to the replication-independent TLR7 response. In addition, RIG-I–mediated type I IFN was limited by the kinase activities of IRAK1 and IRAK4. Such control mechanisms may exist to restrict the production of excessive amounts of type I IFN, which would be detrimental for the host (55). The mechanism by which IRAK1, IRAK4, or both inhibit RIG-I signaling remains to be explored. Our findings may help to explain how IRAK4-deficient patients survive common viral infections (24). The RLR signaling pathway might compensate for the lack of a functional TLR pathway in these patients. Vaccination of humans with YF-17D induces a robust and rapid type I IFN response as well as the expression of a network of ISGs (2931). Our work provides insights into the mechanism by which vaccination with YF-17D does so.

MATERIALS AND METHODS

Biological materials, antibodies, and reagents

The primers, reagents, antibodies, viruses, cell lines, and shRNAs used in this study are summarized in tables S1 to S6, and most have been described previously (3337, 44, 46, 56, 57). Vero and Huh7 cells were generally used at 5 × 105 cells per condition. Blood from healthy adult donors was provided by the Ètablissement Français du Sang, and PBMCs were isolated with Ficoll-Paque (GE Healthcare). pDCs were depleted from intact PBMCs with the CD304 (BDCA-4/Neuropilin-1) MicroBead kit (Miltenyi Biotec). The negative fraction was collected and constituted pDC-depleted PBMCs. Intact and pDC-depleted PBMCs were used at 1 × 105 cells per condition. Alternatively, pDCs were isolated through depletion of non-pDCs with the pDC isolation kit II (Miltenyi Biotec). Isolated pDCs were generally used at 2.5 × 104 cells per condition, with a typical purity of >94%. The linearized YFRP-IRES-Neo plasmid (58) was transcribed in vitro, and 8 × 106 Vero cells were subjected to electroporation with 10 μg of RNA to generate the YFV replicon cell line (YFRP). Twenty-four hours after electroporation, the cells were selected with G418 (400 μg/ml; Invitrogen). YF-17D (table S4) was concentrated by polyethylene glycol 6000 precipitation and purified by centrifugation in a discontinued gradient of sucrose. Sucrose-purified viruses were used for all experiments. YF-17D was inactivated by exposure to UV light (4.75 J/cm2) for 10 min at a distance of 10 cm. The complete loss of detectable infectivity after UV exposure was confirmed by plaque assay. YF-17D was titrated by plaque assay as previously described (59).

Generation of shRNA-expressing Gen2.2 cells by lentiviral transduction

Gen2.2 cells were transduced with lentiviral vector particles expressing the appropriate shRNAs (see table S6), as described previously (18).

Semiquantitative and real-time PCR analyses

Total RNA was extracted from cell cultures or cell-free supernatants with the NucleoSpin RNA II kit (Macherey-Nagel). First-strand complementary DNA synthesis was performed with the RevertAid H Minus M-MuLV Reverse Transcriptase. Quantitative real-time PCR was performed on a real-time PCR system (ABI PRISM 7900HT with SYBR Green PCR Master Mix; Life Technologies). Data were analyzed with the 2−ΔΔCT method (60), with all samples normalized to GAPDH. All experiments were performed in triplicate. Real-time PCR was used to measure viral load. GE concentrations were determined by extrapolation from a standard curve generated from serial dilutions of the plasmid encoding the subgenomic YFV replicon.

Enzyme-linked immunosorbent assay

Cell culture medium was analyzed by ELISA to determine the amounts of secreted IFN-α with the Verikine Human IFNα Multi-Subtype ELISA Kit (PBL Assay Science).

Immunofluorescence

Gen2.2 cells were seeded onto coverslips coated with poly-l-lysine (1 μg/ml), allowed to adhere for 1 hour, and then infected with YF-17D as indicated in the figure legends. Alternatively, Vero cells were seeded onto noncoated coverslips, infected with YF-17D, and then cocultured with CFSE-labeled Gen2.2 cells. Cells were fixed with 4% paraformaldehyde (PFA), permeabilized with 0.5% Triton X-100 in phosphate-buffered saline (PBS), and then blocked with PBS, 0.05% Tween, and 5% bovine serum albumin (BSA) before being incubated with the appropriate primary antibodies. Coverslips were mounted on slides with ProLong Gold Antifade Reagent with DAPI (Life Technologies). Images were acquired with a Zeiss Widefield ApoTome inverted microscope or a Zeiss LSM 700 inverted confocal microscope.

Correlative light and electron microscopy

Vero cells were seeded onto cell culture plates containing a glass insert with a grid and were infected for 16 hours with YF-17D. Gen2.2 cells were then added to the plates, and the cocultures were incubated for a further 24 hours. Analyses were performed as described previously (61).

Flow cytometry

YF-17D–infected cells were detected by staining with anti-NS1 or anti-dsRNA antibodies in PBS, 0.05% saponin, and 1% BSA after fixation in 4% PFA. The pDC population within total PBMCs was revealed by double staining with phycoerythrin-conjugated anti-CD303 (BDCA-2) antibody and fluorescein isothiocyanate–conjugated anti-CD123 antibody in PBS and 1% BSA. Typically, 1 × 105 cells were analyzed for each condition. Samples were analyzed as described previously (18).

Statistical analysis

Statistical analysis was performed with Student’s t test for pairwise comparisons or by one-way analysis of variance when comparing more than two sets of values. Each experiment was performed at least three times, unless otherwise stated. Statistically significant differences are indicated as follows: *P < 0.05, **P < 0.01, and ***P < 0.001; ns, not significant.

SUPPLEMENTARY MATERIALS

www.sciencesignaling.org/cgi/content/full/8/366/ra25/DC1

Fig. S1. Analysis of the depletion or isolation of pDCs from human PBMCs.

Fig. S2. TLR7 agonist–induced type I IFN production by pDCs is efficiently blocked by the specific TLR7 inhibitor IRS661.

Fig. S3. YF-17D replicates in Gen2.2 cells and stimulates the RIG-I signaling pathway.

Fig. S4. YF-17D induces nuclear translocation of IRF3 but not IRF7 in Gen2.2 cells.

Fig. S5. SeV stimulates RIG-I signaling in Gen2.2 cells.

Fig. S6. YF-17D–infected Huh7 cells stimulate IFNA expression in Gen2.2 cells.

Fig. S7. Sensing of YF-17D–infected cells stimulates nuclear translocation of IRF7 but not IRF3 in Gen2.2 cells.

Fig. S8. Infection with YF-17D stimulates nuclear translocation of IRF3 in Vero cells.

Fig. S9. Schematic representation of the proposed YF-17D–mediated signaling pathways in human pDCs.

Table S1. Primers used for semiquantitative and real-time PCR analysis.

Table S2. Reagents.

Table S3. Antibodies.

Table S4. Viruses.

Table S5. Primary cells and cell lines.

Table S6. shRNA sequences for knockdown experiments.

REFERENCES AND NOTES

Acknowledgments: We thank J. Plumas and L. Chaperot for the Gen2.2 cells; D. Garcin for the Sendai virus; E. Meurs for the HCVRP cells; M. Flamand for the NS1 antibodies; R. Kuhn for the 17D replicon constructs; R. Randal for the shRIG-I plasmid; D. Arnoult for the pTBK1 antibodies; and C. Schmitt for helping to perform CLEM experiments. Funding: This work was supported by grants from Agence Nationale pour la Recherche (ANR-12-JSV3-003-01), Ville de Paris EMERGENCES Program, CNRS, and the Pasteur Institute. Author contributions: D.B. and N.J. designed the study, analyzed the data, and wrote the manuscript; D.B. performed the experiments with help from M.C., L.S., and L.C.; and O.S. and P.D. contributed reagents and materials. Competing interests: The authors declare that they have no competing interests.
View Abstract

Stay Connected to Science Signaling

Navigate This Article