Our laboratory sought to examine the role of the endogenous autophagy machinery in host defense against a mammalian virus infection. In earlier reports, autophagy genes limited the spread of cell death during the hypersensitive response in plants infected with tobacco mosaic virus (TMV)1
and protected flies against lethal infection with vesicular stomatitis virus (VSV).2
However, no studies had established a direct role for autophagy in host defense against viral infection in a vertebrate model system. To examine this question, we studied the effect of neuronal inactivation of the Atg5
autophagy gene on the pathogenesis of neonatal CNS infection with Sindbis virus (SIN), an enveloped, positive-strand RNA virus in the alphavirus genus that serves as an animal model for human arthropod-borne encephalitides.3
First, we performed in vitro studies to determine whether SIN induces autophagy, whether SIN is targeted to autophagosomes, and whether autophagy controls SIN replication.3
We found that SIN infection induces autophagy in vitro, as evidenced by the induction of GFP-LC3 puncta, and the degradation of p62 in SIN-infected murine embryonic fibroblasts (MEFs) and mouse neuronal cells. In contrast to the findings of Shelly et al.,2
SIN-induced autophagy requires viral replication, as autophagy is not observed in cells infected with UV-inactivated SIN. We also found that the SIN capsid protein colocalizes with GFP-LC3 in wild-type but not in Atg5-deficient MEFs. This colocalization represents the targeting of SIN capsid protein for autophagic degradation, as demonstrated by electron microscopy (EM) and live cell imaging. Although Atg5 is required for SIN capsid autophagic degradation, it is not required for the control of viral replication, since no differences in viral growth curves are observed in Atg5-deficient versus wild-type MEFs or embryonic stem cells. Thus, in vitro, SIN replication induces an autophagic response that results in the degradation of viral capsid, but not in the control of viral replication.
Next, we used three complementary approaches to inactivate the essential autophagy gene Atg5
in SIN-infected neurons in vivo. In the first model, a double subgenomic SIN vector4
was used to express a dominant-negative mutant version of Atg5 (Atg5K130R
) in infections of wild-type mice. In the second model, the SIN vector was used to express Cre recombinase, resulting in the deletion of floxed alleles of Atg5
only in infected neurons of Atg5flox/flox
mice. In the third model, we infected animals with neuronal specific deletion of Atg5
) with wild-type SIN. In each model, we observed a significant increase in mortality in animals in which Atg5 function was disrupted or Atg5
was deleted, demonstrating that the endogenous autophagy machinery functions to protect against viral infection in vertebrates.
Somewhat surprisingly, but nonetheless consistent with our in vitro findings in Atg5-deficient cells, the increased mortality in mice disrupted of Atg5 is not associated with increased viral titers in the brain. Also, there were no differences observed in the levels of type I interferon, a critical mediator of host defense against SIN infection, in the brains of mice with intact versus those with disrupted neuronal Atg5 function. Yet, there were striking changes noted at the histopathological level. Mice with disrupted neuronal Atg5 have a marked delay in the clearance of SIN antigens, increased accumulation of p62 aggregates, and increased apoptosis of infected neurons. While it is difficult to conclude cause and effect relationships from in vivo viral pathogenesis studies, these results are most consistent with a scenario in which defective cell-autonomous clearance of protein aggregates produced during viral infection results in neuronal death, ultimately leading to decreased survival of the infected animal.
Our in vivo studies led us to ask the question—how are SIN proteins targeted to the autophagosome? As discussed in detail in other articles in this issue of Autophagy
, p62 is an adaptor protein, containing a ubiquitin-binding association (UBA) domain and an LC3-interaction region (LIR), that targets ubiquitinated substrates to the autophagosome. In recent years, the role of p62 in the selective autophagy of cellular proteins and bacteria has received much attention, whereas its potential role in targeting viral proteins had been unexplored. However, in retrospect, a potential link between selective autophagy adaptors and host defense against viruses was suggested decades ago; ref(2)P, the Drosophila melanogaster
p62 ortholog, was found to be responsible for restriction of sigmavirus replication5,6
and ref(2)P was found to bind sigmavirus capsid in coimmunoprecipitation studies.7
In addition, mutations in the UBA domain of p62 are associated with Paget disease of the bone (reviewed in ref. 8
), which has long been suspected to have an underlying viral etiology, as intracellular inclusions resembling paramyxovirus nucleocapsids have been detected in ultrastructural studies of Pagetic bone lesions.9
These reports, coupled with our observations of p62 aggregates in the neurons of SIN-infected mice lacking intact Atg5 function, led us to hypothesize that p62 might target SIN antigens for autophagic clearance. Indeed, we identified SIN capsid as a p62-interacting protein in coimmunoprecipitation studies, and found that siRNA-mediated knockdown of p62 blocks SIN capsid colocalization with GFP-LC3, delays clearance of SIN capsid from infected cells, and accelerates apoptosis of infected cells. Thus, the SIN capsid protein is delivered to the autophagosome by a mechanism involving the p62 adaptor, which helps to protect against virus-induced cell death. Our findings provide the first example of a viral protein serving as a target for selective autophagy, and suggest a role for selective viral autophagy in cell survival during viral infection.