Given the importance of autophagy in antiviral immunity, it is perhaps not surprising that viruses have evolved mechanisms to evade and/or subvert this pathway (reviewed in references 9, 11, 14, 35, 37, 60, 61, and 77). Several members of the herpesvirus family, most notably herpes simplex virus type 1, inhibit autophagy within an infected cell and encode proteins that block and/or target intracellular signaling pathways that regulate autophagy (reviewed in references 60 and 61). However, some viruses not only evade autophagy but also appear to take advantage of the process; several RNA viruses induce autophagy and exploit the pathway during their replication (1, 12, 15, 31, 40, 43, 76, 93, 96). Viruses belonging to the Picornaviridae family and the Nidovirales order replicate their genomes on double-membraned vesicles that resemble autophagosomes; these vesicles are notably smaller in size than cellular autophagosomes and are decorated with proteins derived from the autophagic pathway (19, 21, 31, 37, 67, 68, 71, 92). Viral proteins encoded by poliovirus and equine arterivirus can trigger the formation of these autophagy-like vesicles (79, 80), and the expression of a single poliovirus protein, 2BC, is sufficient to induce lipidation of the host autophagy protein light chain 3 (LC3), encoded by the Atg8 gene (87). Taken together, these studies suggest that some viruses subvert the autophagy pathway to generate double-membraned vesicles that provide a surface for RNA replication (8, 37, 88). In addition, these vesicles may permit newly formed virions to escape from infected cells via a nonlytic route (36, 85).

Although studies have demonstrated that the autophagic pathway may play an important role in virus infection in vitro, either to promote or to restrict viral replication, we are just beginning to appreciate and understand the function and effects of autophagy for virus infections in vivo. Autophagy acts in an antiviral fashion to limit tobacco mosaic virus replication and programmed cell death in plants (46), to prevent a pathogenic infection with vesicular stomatitis virus in flies (73), and to protect against fatal encephalitis in Sindbis virus- or herpes simplex virus type 1-infected mice (45, 59, 63). Nonetheless, to date there is a dearth of in vivo studies; animal models of virus infection are needed in order to better define the antiviral role of autophagy in vivo (41, 62). In addition, studies that address the role of viral subversion of autophagy in vivo are warranted. Does this process occur within infected animals, and is it required for viral replication in particular cell types or for viral pathogenesis? Recent studies have shown that autophagy not only promotes the replication of hepatitis B virus and enterovirus 71 in vitro but also may be induced by infection in vivo, potentially to benefit the virus rather than the host (28, 78).

Type B coxsackieviruses (CVBs) are members of the Picornaviridae family and Enterovirus genus and, as such, are closely related to polioviruses. CVBs are important human pathogens that often induce severe acute and chronic diseases and cause morbidity and mortality (69, 91). CVBs are the most common cause of infectious myocarditis (38, 82) and frequently trigger pancreatitis and aseptic meningitis (7, 16, 29, 51). Tissue culture studies (93) have shown that CVB type 3 (CVB3) promotes LC3 conversion and autophagosome accumulation in virus-infected cells in vitro and that modulation of the autophagic pathway (using chemicals or small interfering RNA-mediated knockdown) to enhance or dampen autophagy results in an increase and a decrease, respectively, in viral protein expression and/or viral titers; however, the reported changes in viral titers were modest (2- to 4-fold). In the present study, we examine whether CVB3 activates the autophagic pathway in vivo, specifically in pancreatic acinar cells, which are a natural primary target for this virus. Using a mouse model of CVB3 infection, which faithfully recapitulates most aspects of CVB disease in humans, we demonstrate that this virus triggers LC3 conversion and also modulates other components of the autophagy machinery. In addition, using a recombinant CVB3 (rCVB3) that expresses Discosoma sp. red fluorescent protein (DsRed-CVB3), we identify virus-infected cells in situ and show that CVB3 infection increases autophagosome abundance in vivo. Lysosomal-associated membrane protein 1 (LAMP-1) immunostaining confirmed that amphisomes are generated in virus-infected cells but that autophagic flux was not substantially enhanced as the infection progressed; rather, there appears to be a substantial blockade in fusion with lysosomes. Finally, transmission electron microscopy (TEM) ultrastructural analysis of the infected pancreas confirmed that double-membraned autophagy-like vesicles as well as very large autophagic compartments (for which we have coined the term “megaphagosomes”) were generated in acinar cells following virus infection. Overall, these data provide compelling evidence that CVB3 induces autophagy in vivo and suggest that this picornavirus may subvert this process in a mammalian host.

In this study, we evaluated three distinct criteria in order to determine if CVB3 induces autophagy in pancreatic acinar cells. First, we sought a biochemical signature of autophagy induction by analyzing the conversion of LC3-I to LC3-II by Western blotting. The level of LC3-II increased significantly on day 2, and by day 3 nearly all of the endogenous LC3 present in pancreas was converted; this was accompanied by a loss of the monomeric and cleaved forms of Atg5; only the Atg5-Atg12 complex remained (Fig. 1A and B). Second, we assessed autophagic flux by monitoring, over the course of infection, the levels of p62 and fragmented GFP. The substantial increase in the autophagy substrate p62 by day 2 to 3 p.i. (Fig. ​(Fig.2B)2B) suggests that autophagic flux is markedly reduced. Others have reported that p62 levels were unchanged in CVB-infected tissue culture cells (93); the reason for this apparent discrepancy is unclear but may reflect major differences between the experimental systems (tissue culture cells versus pancreatic tissue in vivo). We also found that the amount of cleaved GFP increased significantly by day 3 p.i. (see Fig. S2 in the supplemental material), but this is difficult to interpret for the reasons described above. Third, we performed high-resolution confocal imaging of vibratome-cut pancreas sections to quantitate the number of GFP-LC3⁺ punctae in uninfected and virus-infected tissue. DsRed-CVB3 triggered not only a 2-fold increase in the number of autophagosomes in strongly infected (DsRed bright) acinar cells but also an ~3-fold increase in their average surface area (Fig. ​(Fig.4B).4B). These data are consistent with the hypothesis that CVB3 recruits autophagosomes into virus factories to generate a scaffold for the replication complex.

Others have found that poliovirus RNA polymerase is associated with membranes isolated from infected tissue culture cells and that the purified protein can form a lattice in vitro (48); the authors proposed that poliovirus replication in vivo might take place on a “shell” of membrane-associated polymerases. The highly organized structure shown in Fig. ​Fig.6L6L appears to be a paracrystalline lattice formed by individual components, each ~8 nm in diameter. CVB RNA polymerase (22) is structurally very similar to poliovirus RNA polymerase, a single molecule of which is ~8.8 by 8.8 by 15.8 nm in size (24). We suggest that the lattice shown in Fig. ​Fig.6L6L represents an array of CVB RNA polymerases within an infected pancreatic acinar cell and speculate that CVB replication in vivo may take place on this sheet of polymerases, consistent with the aforementioned proposal.

In principle, the accumulation of autophagosomes that we observed in CVB-infected acinar cells could result from an increase in autophagosome formation, from a decrease in their degradation, or both. Previous studies in vitro have suggested that poliovirus and CVB3 can inhibit the maturation/degradation of virus-induced double-membraned vesicles (or autophagosomes) by blocking their fusion with lysosomes (86, 93). Indeed, coexpression of the poliovirus proteins 2BC and 3A is sufficient to induce the colocalization of LAMP-1 (a marker of the endolysosomal compartment) with GFP-LC3 (31). We therefore examined whether autophagosomes in CVB3-infected acinar cells were able to fuse with endosomes/lysosomes or if this process was inhibited by the virus in vivo. Many GFP⁺ vesicles in DsRed-CVB3-infected cells also contained LAMP-1 (Fig. ​(Fig.5),5), suggesting that amphisomes are generated; thus, CVB3 does not appear to prevent autophagosome-to-endosome fusion. It is more difficult to evaluate the effect (if any) of CVB3 on subsequent steps, i.e., the formation of autolysosomes. However, megaphagosomes have not been observed during the upregulation of autophagy that takes place in response to normal stimuli, such as food restriction, and their presence in CVB-infected cells therefore suggests that the virus may interrupt autophagic processing at some stage of the pathway. These very large vesicles, which were sometimes seen in clusters, appear to be related to autophagy because they were usually GFP⁺ by confocal microscopy (Fig. ​(Fig.44 and ​and5),5), and some also scored positive by immuno-EM (Fig. ​(Fig.7).7). Mice deficient in LAMP-2, a key constituent of the lysosomal membrane, develop large autophagic vacuoles in several cell types, including pancreatic acinar cells, and the authors suggested that these large vacuoles resulted from failure of fusion of autophagosomes with endosomes/lysosomes (83). Giant perinuclear autophagosomes have also been observed in cells from mice with nuclear envelopathies and in A549 human lung epithelial cells infected with influenza A virus (17, 49). For both of these models, it was suggested that clusters of smaller autophagosomes might fuse together to form a giant autophagosome; the influenza A virus M2 protein may play a role in this process by preventing autophagosome maturation (17, 66, 70). Likewise, CMVs were abundant in wt CVB3-infected acinar cells (Fig. ​(Fig.6),6), and small autophagosomes were often seen close to the membrane of the giant autophagosomes and/or were contained within the larger structures (Fig. 4D and E). Therefore, we suggest that CVB3 infection (i) increases the rate at which new autophagosomes are formed and (ii) inhibits their progress along the autophagy pathway. Consequently, (iii) these abundant smaller vesicles coalesce to form megaphagosomes. Where might CVB inhibit autophagy flux? The megaphagosomes usually express the endosomal/lysosomal marker LAMP-1 (Fig. ​(Fig.5),5), suggesting that they are very large amphisomes. These GFP⁺/LAMP-1⁺ vesicles are less likely to be autolysosomes because, as noted above (and see Fig. ​Fig.2A),2A), autolysosomes in GFP-LC3 mice should be LAMP-1⁺ but, because of their acidic environment and protease content, should show minimal GFP fluorescence. Thus, we suggest that CVB3 permits amphisome formation but limits the subsequent fusion of amphisomes with lysosomes. This is consistent with the observation that p62 accumulates in infected tissue by day 2 to 3 p.i. (Fig. ​(Fig.2B).2B). Nevertheless, this inhibition, although profound, may be incomplete: we occasionally observed large vesicles that were predominantly LAMP-1⁺ with minimal GFP fluorescence (Fig. ​(Fig.5).5). Normal endosomes and lysosomes would be expected to show the same fluorescent qualities, but the unusually large size of these vesicles suggests that they may have been generated from megaphagosomes that have escaped the effects of CVB, becoming autolysosomes. In summary, CVB3 appears to block autophagosome fusion with lysosomes in vivo, as has been proposed for poliovirus in tissue culture cells (86).

What advantages might accrue to RNA viruses that increase autophagosome abundance and also prevent their maturation into autolysosomes? This strategy may benefit the virus in several ways: (i) by inducing the accumulation of intracellular vesicles on which the viral replication machinery is assembled; (ii) by preventing degradation of newly formed virions inside autolysosomes; and (iii) by limiting the interaction between viral RNA and the single-stranded sensor Toll-like receptor 7 (TLR-7), which is present in late endosomes and lysosomes and which is thought to play a role in innate responses to CVB. By terminating the autophagy pathway at the stage of amphisomes, CVB ensures that its RNA is separated from TLR-7 by the intact inner lipid bilayer that is derived solely from the TLR-deficient autophagosome; only upon degradation of this inner membrane (in an autolysosome) will the viral genome be introduced to TLR-7. Furthermore, (iv) others have proposed that the inhibition of autophagic flux and the generation of autophagy-like vesicles may permit the extracellular release of virus without cell lysis (6, 88). Finally, (v) autophagy can be important in antigen presentation via MHC class I (13) and MHC class II (42, 64); we speculate that by blocking the autophagy pathway, CVB3 infection may compromise the host's capacity to mount strong T-cell responses, possibly contributing to the weakness of the CVB-specific T-cell responses that this laboratory has recently reported (34).