The innate immune system is considered as the first line of defense against invading pathogens like viruses and bacteria. Upon viral infection, many cytosolic and endosomal immune sensors detect different components of viral nucleic acids to elicit an innate immune response. There are three classes of PRRs or immune sensors: Toll-like receptors (TLRs), retinoic acid-inducible gene I (RIG-I)-like receptors (RLRs) and nucleotide oligomerization domain (NOD)-like receptors (NLRs) [52]. TLRs are located either on the plasma membrane or within endosomal compartments. In contrast, RLRs and NLRs are mainly localized in the cytoplasm. Among the TLRs, TLR3, 7/8 and 9, which are mainly localized in endosomal compartments, recognize dsRNA, ssRNA and CpG DNA, respectively [52]. Many RNA viruses release their viral RNA genome into the cytoplasm to establish sites of RNA replication, thus allowing cytosolic PRRs to be activated [53]. However, activation of endosomal TLRs can be limited, because viral RNA is packed into the nucleocapsid or sequestered in membrane-associated replication complexes and thus, well protected from the endosomal environment. Nonetheless, TLRs frequently trigger an immune response targeting a number of viral infections via poorly defined mechanisms [54].

The first report demonstrating that autophagy can be involved in delivering PAMPs or viral nucleic acids to endosomal TLRs comes from the study of VSV-infected plasmacytoid dendritic cells (pDCs) [55] (Figure 3). The authors showed that the production of interferon-α (IFN-α) by VSV-infected pDCs depended on the autophagic delivery of viral replication intermediates to TLR7 present on late endosomes. This activation of IFN-α production required active viral replication. Furthermore, upon treatment of VSV-infected pDCs with Wortmannin or 3-methyladenine (3-MA) that both inhibit Vps34 activity, IFN-α production was blocked. Finally, the authors showed that Atg5-/- mice exhibited a similar phenotype and carried higher viral loads, suggesting that autophagy is important in controlling VSV replication in vivo. However, the exact mechanism by which viral PAMPs are selectively targeted by autophagosomes to the TLR-resident late endosomal compartment remains to be determined.

The involvement of autophagy in delivering viral PAMPs to endosomal PRRs is also conserved in non-immune cells. Studies with Coxsackie virus B3 (CVB3)-infected kidney fibroblasts demonstrated autophagy-dependent activation of TLR3 [56]. The same study revealed that TLR3-decorated vesicles colocalized with LC3-positive structures and that the activation of TLR3 via poly (I:C) or CVB3 infection was sensitive to inhibitors of autophagy induction, autolysosomal acidification and lysosomal proteases. Furthermore, even in the absence of poly (I:C) or CVB3 infection, TLR3 still localized to LC3-positive structures, arguing that even under basal conditions, the autophagosome might fuse with TLR3-positive endosomal vesicles [56]. The resulting ‘mixed’ compartments might therefore serve as antiviral platforms specialized in sensing viral PAMPs and eventually triggering downstream signaling events.

Induction of autophagy can also occur via direct activation of TLRs [57]. Upon stimulation of TLR 3, 4 and 7 with different purified PAMPs, autophagy was shown to be induced in a MyD88- and/or TRIF-dependent manner [58,59]. MyD88 and TRIF function as adaptors through which TLR signaling is mediated. Upon TLR activation, MyD88 and TRIF directly interact with Beclin-1 leading to the formation of Beclin-1-Vsp34 complexes that induce autophagy [59].

The role of autophagy in delivering viral PAMPs to TLRs and the subsequent induction of autophagy upon TLR activation indicates the existence of a positive feedback loop. This feedback loop might not only play an important role in the autophagic sequestration and delivery of PAMPs to endosomal compartments, but also serve for autolysosomal degradation of virions and viral proteins, as well as for antigen presentation via MHC-I and MHC-II molecules, thereby triggering the adaptive immune response.

Autophagy adaptors, referred to as SLRs, are considered as PRRs that can selectively target a variety of pathogens for autophagic degradation [60]. The cytoplasm offers a diverse range of autophagic targets that vary in size and complexity, ranging from protein aggregates up to complete organelles, which can be selectively recognized and sequestered by proteins that function as autophagic adaptors [60]. The main autophagic adaptors that are classified as SLRs in response to bacterial and viral infections, include p62, NBR1 (neighbor of BRCA1 gene 1), NDP52 (nuclear dot protein 52 kDa) and optineurin [60]. Typically, SLRs contain cargo recognition and capture domains, LC3-interacting regions (LIR) to target captured cargo (bacteria or viruses) to the autophagosomal compartment and additional protein interaction domains that are involved in inflammatory processes [60]. However, in the case of Salmonella infection, it was demonstrated that pathogen recognition and capture occurs via multiple SLRs, such as p62, NDP52 and optineurin, by recognition of conventional or branched ubiquitin chains that were associated with or in close proximity of cytosolic salmonellae [60,61].

The role of SLRs in the selective autophagic degradation of viruses was recently studied in the context of Sindbis virus infection [37]. Orvedahl and co-workers found that autophagic degradation of Sindbis virus involves the selective degradation of the viral capsid in a p62-dependent manner (Figure 3). The precise signal for capsid recognition by p62 is unclear, but is independent of ubiquitin association, because neurons of infected wild-type or Atg5-/- mice did not contain accumulations of ubiquitinated cellular or viral proteins [37]. Moreover, in an effort to identify host cell factors that target Sindbis virus and HSV-1 for autophagic degradation, the authors performed a genome-wide RNA interference screen. Surprisingly, this screen identified common molecular determinants involved in the targeting of viral nucleocapsids and mitochondria for autophagic degradation [62]. This virophagy and mitophagy is critically determined by SMURF1, an E3 ubiquitin ligase that acts as a bona fide mediator of selective autophagy [62]. SMURF1 might therefore be considered as a novel SLR specifically involved in controlling viral infections. Whether SMURF1 is also involved in autophagy of bacterial invaders remains to be determined.

The role of autophagy is not limited to the innate immune response, but also plays a major role in the adaptive immune response by processing and delivering antigens for presentation in complex with MHC-I and MHC-II molecules [63] (Figure 3). Typically, MHC-I molecules present endogenous antigens that are degraded by the proteasome and transported into the ER, where they are loaded onto MHC-I molecules. After transport to the cell surface, peptide-loaded MHC-I stimulates a CD8+ T-cell response responsible for controlling infections [64]. MHC-II molecules present exogenous antigenic peptides that are processed by lysosomal degradation. Peptides are loaded onto MHC-II and transported to the cell surface to stimulate a CD4+ T- cell response [64]. Notably, exogenous antigens can also be presented on MHC-I molecules to mediate a CD8+ T-cell response by a mechanism termed “cross-presentation” [65]. This is found primarily with dendritic cells, macrophages and B lymphocytes and plays a major role in the immune defense against numerous viral and bacterial infections.

A number of studies have convincingly demonstrated that autophagy can influence the presentation of viral antigens on MHC-I and MHC-II [66]. Studies with HSV-1 showed that at early time points after infection, autophagy is not required for viral antigen processing and presentation of MHC-I peptides, but comes into play at late stage of infection [67]. However, it remains unknown how degradation of HSV-1 particles by the autophagosome results in the presentation of viral antigens to ER localized MHC-I molecules. In the case of Epstein Barr virus (EBV), the viral protein EBNA1 is targeted to MHC-II compartments via autophagic uptake and fusion of autophagosomes with these compartments [68]. Inhibition of autolysosomal acidification resulted in enrichment of EBNA1-positive autophagosomes, and knock-down of Atg12 rendered the cells unable to stimulate MHC-II-dependent activation of T-cells [69]. Importantly, already under normal conditions and in the absence of viral infection, autophagy was required for the consecutive delivery of antigens to MHC-II molecules, arguing for a crucial role of autophagy in MHC-II presentation of antigens in vivo [70]. Furthermore, by fusing the matrix protein of influenza virus to LC3, an increased MHC-II activation of CD4+ T-cells was observed, thus revealing an unexplored strategy for vaccine development [70].

Most positive-strand RNA viruses re-shape the endomembrane system in order to create membrane-associated replication ‘factories’ [71]. This reorganization of intracellular membranes frequently requires autophagy and was first described for the poliovirus [72]. By using thin section EM analysis of infected cells, double-membrane vesicles (DMVs) were detected that are characteristic of autophagosome-like structures (Figure 4). These structures stained positive for the viral protein 2BC, and they did not co-fractionate with markers of other known organelles, highlighting their distinct nature [72]. Expression of the viral proteins 2BC and 3A resulted in the lipidation of LC3 and induced the formation of autophagosome-like DMVs [72]. These properties are not unique for poliovirus, but have been found for several other picornaviruses such as CVB3, enterovirus 71 (EV71) and foot-and-mouth disease virus (FMDV) [42,73,74]. Cells infected with these viruses also contain accumulations of DMVs that are not only morphologically similar, but also contain LC3 as well as viral replicase proteins (Figure 4). However, it is important to note that the membrane composition of picornaviral replication sites is heterogeneous, and other mechanisms have been suggested to explain their formation and composition [75,76]. Importantly, key elements of the secretory pathway are involved in the biogenesis of these membranous structures, including COP-II coated vesicles, Golgi-specific brefeldin A-resistant guanine nucleotide exchange factor-1 (GBF-1), ADP-ribosylation factor 1 (ARF1) and phosphatidylinositol 4-kinase III beta (PI4KIIIβ) [77]. Finally, it was also shown that poliovirus, EV71 and FMDV can replicate in the absence of autophagy, although in some cases virus production was decreased [18,42,73,78]. These studies suggest that autophagosomes, per se, are dispensable for the biogenesis of viral replication sites, although some autophagy proteins can contribute to their morphological heterogeneity.

Conflicting reports have been published as to the role of autophagy for the HCV replication cycle [23,26,79,80]. For instance, Sir and co-workers used the highly permissive cell clone Huh 7.5 and found that transfection of HCV RNA induces autophagic vacuoles and leads to an incomplete autophagic process that promotes viral replication [23]. Dreux and co-workers employed Huh7 cells infected with HCV and found that autophagy was only required for initiating the translation of incoming viral RNA, but not for establishment or maintenance of viral replication [26]. Moreover, in Huh7 cells containing HCV replicons of different genotypes (1a, 1b or 2a), the knock down of Atg proteins did not affect expression of viral proteins [81]. In contrast, in another study that employed Huh7.5 cells, the knock-down of Atg7 dramatically decreased production of HCV particles, but did not affect viral replication [80]. Moreover, a recent report showed that knock-down of the autophagy-related proteins beclin 1 (BCN1) and Atg7 inhibited HCV replication in immortalized human hepatocytes (IHHs) [82]. These conflicting results might be due to the use of different cell lines. In fact, a recent study highlighted the importance of the UPR and autophagy during HCV infection and demonstrated that autophagy enhances HCV replication by inhibiting an innate immune response [36]. Since Huh7 clones differ in their capability to mount an innate antiviral defense, the use of different cell lines could explain, at least in part, the contradictory results.

More recently, Sir et al proposed a direct role of autophagy in serving as platforms for viral RNA replication [79]. In this study, the authors describe the localization of HCV replicase proteins NS5A and NS5B as well as nascent HCV RNA to autophagosome-like structures. However, these data are in direct opposition to studies showing that HCV proteins do not colocalize with autophagosomes in infected cells [23,49,80]. Thus, although autophagy is required for HCV replication, the precise role of autophagy or autophagy-related proteins in viral replication remains to be determined.

In the case of coronaviruses, autophagosomes were initially proposed to be sites of RNA replication, since viral replicase proteins colocalized with LC3 and the deletion of Atg5 restricted viral replication [83]. However, subsequent studies suggested that the role of autophagy in viral replication might be a cell type-specific effect [84]. Electron tomography and 3D reconstructions of coronavirus-infected cells demonstrated that DMVs form an adjoining network likely derived from the ER [85]. Moreover, LC3 was localized to the viral replicase machinery, and surprisingly, only unlipidated LC3-I, but not LC3-II, decorated the replication machinery, as well as replication membranes [86]. Furthermore, viral DMVs also stained positive for ER-associated degradation (ERAD) markers. This data suggests that coronaviruses do not require autophagosomes, per se, but utilize components of the autophagy pathway. Another example of viruses utilizing autophagy for viral replication is rotavirus. It was found that the viral protein NS4P, a marker of rotavirus replication sites, colocalized with LC3-positive structures [19] (Figure 4). These structures did not stain positive for lamp1, suggesting a block in fusion of virus-associated autophagic structures with lysosomal compartments [19]. However, the precise role of autophagy in viral replication remains to be explored.

Apart from a direct contribution to the formation of membranous replication factories, autophagy can promote viral replication in other ways. For instance, a recent study by Heaton and co-workers identified a novel proviral mechanism of autophagy: the production of energy [87]. Taking into account the so-called lipophagy, i.e., the degradation of triglycerides stored in lipid droplets (LDs), the authors demonstrated that DENV infection of hepatocytes induces autophagy-mediated LD degradation. This results in a mobilization of triglycerides that are used for ATP production via β-oxidation. The signaling pathway controlling the selective degradation of LDs via autophagy and the mechanism that regulates ATP production are not clear. Nevertheless, this study is the first to link of viral infection to autophagy-dependent regulation of lipid metabolism. It will be interesting to determine whether this link is unique to DENV or applies to other (LD-dependent) viruses.

Upon replication of the viral genome, it has to be packaged into a virus particle (assembly) that is released from the infected cell (exit). For many viruses, autophagy has been reported to regulate also these late steps of the viral replication cycle [18,28,78]. For instance, in the case of hepatitis B virus (HBV), efficient envelopment of viral particles was shown to depend on induction of autophagy [88] (Figure 4). Cells that are deficient in autophagy produce significantly lower amounts of nucleic acid-containing intra- and extracellular enveloped particles. Furthermore, the major HBV envelope protein (HBsAg) was shown to bind to and colocalize with LC3-I and LC3-II during HBV infection or upon ectopic expression of HBsAg, indicating that this interaction might be important for acquiring the viral envelope. Autophagosomes might serve as membrane source; alternatively, intermediate compartments might be used to transport HBV to the site of envelopment, which is thought to occur at a post-ER/ pre-Golgi site [88].

Upon infection of macrophages with HIV-1, the viral assembly protein Gag interacts with LC3-II, and this association allows the processing of the Gag subunit p24. Confocal microscopy and EM analyses revealed that gag-derived matrix (MA) protein p17 (a component of HIV-1 particles) colocalized with membrane compartments that stained positive for LC3 and that were in close proximity to the plasma membrane [28] (Figure 4). These structures are similar to those proposed previously as sites of HIV-1 assembly and budding in macrophages [28]. Furthermore, the ability of the HIV-1 Nef protein to inhibit autophagosome acidification was also shown to be required for release of infectious virions [28]. How acidification affects Gag-LC3 interaction still remains to be addressed.

Lastly, autophagy has been shown to serve in the non-lytic release of non-enveloped picornaviruses [18,73] (Figure 4). Pharmacological inhibition of autophagy or silencing of expression of key autophagy factors results in a decrease in infectious particle production. Moreover, disrupting the microtubule network by using nocodazole treatment or challenging cells with a mutant virus resulted in increased mobility of poliovirus-induced GFP-LC3 structures, concomitant with an increase of extracellular infectious particles [78]. Although the mechanism of this non-lytic virus release is not known, recent studies have confirmed the role of autophagy in the unconventional secretion of diverse substrates [89].

Several recent studies have highlighted how viruses manipulate autophagy to dampen key molecules of the innate as well as the inflammatory (NF-kB-dependent) immune response. In an elegant study by Fliss and co-workers, murine cytomegalovirus (MCMV) infection-induced autophagy was shown to selectively target NEMO (NF-kB essential modulator) to degradation in lysosomes, thus inhibiting an inflammatory response [90] (Figure 4). The authors show that inhibition occurs via an interaction of the viral protein M45 with NEMO that is required to selectively relocalize NEMO to GFP-LC3-positive autolysosomal structures [90].

A recent report by Gre´goire et al. showed that the IRGM protein (immunity-associated GTPase family M) is a common target of several RNA viruses to subvert the autophagy network [91]. By using a yeast two-hybrid screening approach and bioinformatic analyses. the authors determined the ability of 83 proteins from different RNA virus families to interact with 44 human autophagy-associated proteins. It turned out that the autophagy network is highly targeted by RNA viruses. Importantly, IRGM, the most frequently targeted protein, interacted with the autophagy proteins Atg5, Atg10, MAP1CL3C and SH3GLB1 (SH3-domain GRB2-like endophilin B1). Silencing of IRGM expression impaired both Measles virus (MeV), HCV and HIV-1 induced autophagy and virion production. Furthermore, IRGM interacted with MeV capsid, HCV NS3 and HIV Nef. These interactions were sufficient to induce autophagy via an IRGM-dependent pathway. This work revealed a primary role of IRGM in virus-induced autophagy and identified a large range of RNA viruses that use common strategies to manipulate autophagy in order to promote viral replication [91].

As alluded to in the previous section, the direct contribution of autophagy to promoting HCV replication is discussed controversially. However, two recent studies provided insights as to how autophagy might promote HCV replication indirectly, namely by blocking the innate immune response. The study by Ke and Chen demonstrated that UPR induced by HCV activates the complete autophagy pathway, which is required for efficient viral replication [36]. Importantly, this enhancement of replication correlated with suppression of innate antiviral immunity. Interestingly, the analogous activation of UPR and inhibition of IFN-β induction was found with a PAMP derived from DENV, arguing that both viruses utilize the same strategy to maximize RNA replication and to suppress the innate antiviral defense (Figure 4). In agreement with these conclusions, Shrivastava and coworkers found that gene silencing of Beclin 1 and Atg7 in immortalized human hepatocytes not only reduced HCV RNA replication, but also resulted in the activation of IFN signaling [82]. These studies reveal a novel mechanism employed by HCV and DENV to evade an anti-viral innate immune response.