Our study delineates some features of HIV innate recognition. Cell-free HIV particles are generally considered as poor inducers of type I IFN 
. We show here that HIV-infected cells are much more potent stimulators of IFN than free viral particles, confirming and extending previous findings 
. Various cell types may recognize HIV-infected lymphocytes. Among PBMCs, pDCs are the main cells detecting HIV-infected cells, since their removal strongly decreases IFN release. TLR7 senses HIV components, but additional pathways are operative. Viral fusion is an important step promoting HIV detection, whereas reverse transcription and subsequent events of the viral life cycle are not necessary to activate an immune response. Using a model of cells lacking TLR7, we demonstrate that the IRF3 pathway participates in HIV recognition. Therefore, the host innate response to HIV involves the integration of multiple sensor pathways.
Why detection of HIV-infected cells is more efficient than that of cell-free virions? A contact with infected cells promotes a massive and rapid transfer of viral material to target cells 
, as visualized here in pDCs by immunofluorescence and flow cytometry. We show here, by shaking cocultures and by using transwell chambers, that a direct contact between infected cells and recipients is required to trigger high levels of IFN release by these latter. Various modes of cell-to-cell HIV transfer have been reported in culture, including synapses, polysynapses, filopodial bridges and nanotube-like structures 
. HIV dissemination involves viral endocytosis in target cells 
. Whatever the mode of cell-to-cell transfer, the viral influx is quantitatively more important than capture of cell-free viral particles and may explain in part the potent induction of IFN production. Furthermore, infected cells may produce additional cellular or viral components that are not found in virions, which may enhance the activation of target cells.
Detection of HIV-infected cells by pDCs and PBMCs requires functional viral envelope glycoproteins. Envelope-deleted viruses are not or poorly recognized, mainly because target cells capture limited amounts of viral material in the absence of Env glycoproteins. When produced in MT4C5 cells, the fusion-defective envelope mutant (F522Y) is efficiently captured by PBMCs, but its ability to stimulate pDCs is reduced. This raises interesting mechanistic insights. F522Y HIV is normally up-taken in the endosomal compartment but further access to the cytosol is compromised 
. Thus, there are multiple sites of recognition of incoming material from HIV-infected cells. In endosomes, TLRs may detect incoming HIV, as illustrated by the inhibitory effect of Bafilmoycin A1 and of TLR antagonists. After fusion, additional mechanisms of detection may enhance IFN production.
Several lines of evidence suggest that viral replication and productive infection of PBMCs are not necessary to trigger a response to infected cells. Nevirapine, a reverse transcriptase inhibitor, does not inhibit IFN release by PBMCs. However, it is possible that Nevirapine does not fully inhibit viral DNA synthesis in coculture systems of viral cell-to-cell spread. As incomplete viral DNA products represent a source of cytoplasmic nucleic acids that activate an innate immune response 
, we examined the activity of a series of viral mutants, arrested at different steps of the viral cycle. We show that HeLa cells expressing defective viruses, inactivated either in their reverse transcriptase, RNAse H, integrase, protease or nucleocapsid, stimulate IFN release in PBMC cocultures as efficiently as WT HIV. In this coculture system, in which HeLa cells produce relatively low levels of virus, viral fusion activity enhances recognition, as demonstrated with ΔEnv and F522Y mutants that triggered low levels of IFN production. Whether virus-cell or cell-cell fusion, or a combination of both, cause PBMCs activation will require further investigation. Cell-cell fusion is a feature of many viral infections, including HIV. The resulting syncytia may represent a structure producing high levels of IFN, as reported for Measles-induced giant cells 
. Noteworthy, Env expression alone was not sufficient to activate PBMCs. It will be worth assessing the stimulating effect of donor cells producing particles devoid of HIV RNA.
The majority of HIV particles produced by an infected cell are defective or lead to abortive infection  
. Defective viruses are generated by reverse transcriptase errors, editing activity of APOBEC proteins, and by various other mechanisms. Our results suggest that cells producing non-replicating viruses, if retaining fusion potential, may represent an underestimated source of innate immune activators. Our results also confirm that abortive infection may be sensed in target cells 
, thus triggering apoptotic and inflammatory events.
To gain a mechanistic insight into how pDCs recognize HIV-infected cells, we used the Gen2.2 pDC-like cell line as a model 
. We show that Gen2.2 cells, like primary pDCs, are sensitive to HIV replication and release IFN upon coculture with HIV-infected lymphocytes. We generated TLR7-negative Gen2.2 cells by RNA silencing. These cells were impaired in IFN production when cocultivated with HIV-infected lymphocytes, whereas CpG-induced TLR9 signaling was normal. These results provide a direct demonstration that TLR7 mediates recognition of HIV-infected cells. In addition, Gen2.2 cells recognized fusion-defective HIV significantly less potently than the wild-type virus, confirming that sensing of HIV-infected cells may occur at different intracellular locations in pDCs.
In this study, we also used 293T-derivatives as target cells. These non-hematopoïetic cells are not natural targets of HIV infection, but represent a valuable model to study HIV recognition in cells devoid of TLR7. In these targets, activation of the IFNβ promoter occurred upon coculture with HIV-infected lymphocytes, and not with free virions. Signaling required the expression of fusogenic envelope in donor cells, and the presence of cognate receptors (CD4 and CXCR4) in target cells. This strongly suggests that viral or cellular components originating from donor cells need to gain access to the cytoplasm of targets to activate the IFNβ promoter. We dissected further the signaling cascade leading to this activation. Promoter activation required the IRF3 transcription factor 
, as demonstrated by silencing this molecule. The role of IRF3 was confirmed by directly visualizing with the ImageStream apparatus its nuclear translocation, when target cells encounter HIV-infected cells. Of note, SeV activated more efficiently the IFNβ promoter than HIV-infected cells. This activation was particularly sensitive to IRF3 signaling (90% decrease in IFNβ promoter activity), and was associated with a marked nuclear translocation of IRF3. Thus, the stimulating effects of SeV and of HIV-infected cells on the IFN pathway are in part similar, rely on IRF3, but are not totally overlapping. That sensing of HIV-infected cells occurs via IRF3 signaling is in line with two recent reports, demonstrating that this transcription factor mediates recognition of HIV particles in different TLR7- cell types, including MDDCs and macrophages  
Our observations raise intriguing questions about the nature of the cytosolic molecules recognizing HIV-infected cells. The cytosolic RLR (RIG-I-like receptors) helicases RIG-I and MDA5 may be involved in this phenomenon, since they directly recognize multiple and distinct forms of intracellular dsRNA. For instance, putative RIG-I ligands include 5-PPP-bearing RNAs, as well as RNAs with complex secondary structures 
. MDA5 binds to long dsRNAs 
. Addressing the role of RIG-I and MDA5 on recognition of HIV-infected cells will require further investigation. It will be also worth examining the implication of other cytosolic sensors that have been recently identified as mediating or modulating recognition of cell-free virions. This includes 3′ repair exonuclease 1 (TREX1) 
and cyclophilin 
. TREX1 acts as a negative regulator of the IFN response to DNA species generated during reverse transcription of endogenous retroelements 
. TREX1 degrades HIV DNA and has been proposed to be used by the virus to avoid triggering detection by DNA sensors 
. Cyclophilin binding to newly synthesized HIV Gag proteins also triggers an antiviral state in MDDCs 
. It will be worth determining the role of TRIM5 molecules that bind incoming HIV capsids, regulate viral uncoating and thus exposure of HIV nucleic acids in infected cells. Whatever the underlying mechanism, our results demonstrate the existence of TLR7-independent pathways involved in the innate host response against HIV-infected cells.
What may be the relevance of this TLR7-independent signaling pathway? MDDCs are poorly responsive to cell-free HIV: they do not mature nor produce IFN when exposed to cell-free viral particles 
. However, circumventing the restriction of MDDC to infection significantly enhances viral detection 
. This innate response is dependent on cyclophilin A and subsequent IRF3 activation 
. It will be worth determining if HIV-infected cells are sensed by MDDCs more potently than free virions, and if so, examine the role of Gag- cyclophilin interactions in this process.
In sum, our results show that the innate immune system efficiently senses HIV-infected lymphocytes. The potent recognition of HIV-infected cells may overcome the poor detection of cell-free virions. Various cellular sensors are involved in this recognition, which likely represents a driving force regulating the pathology of infection, as well as the generation of an adaptative immune response. How do the different sensor pathways interact to coordinate the host response to viral infection remains an outstanding question. Deciphering the basic mechanisms of HIV recognition has obviously important implications for the development of tailored vaccine strategies.