CAPRIN1 Is Required for Control of Viral Replication Complexes by Interferon Gamma

ABSTRACT Replication complexes (RCs), formed by positive-strand (+) RNA viruses through rearrangements of host endomembranes, protect their replicating RNA from host innate immune defenses. We have shown that two evolutionarily conserved defense systems, autophagy and interferon (IFN), target viral RCs and inhibit viral replication collaboratively. However, the mechanism by which autophagy proteins target viral RCs and the role of IFN-inducible GTPases in the disruption of RCs remains poorly understood. Here, using murine norovirus (MNV) as a model (+) RNA virus, we show that the guanylate binding protein 1 (GBP1) is the human GTPase responsible for inhibiting RCs. Furthermore, we found that ATG16L1 mediates the LC3 targeting of MNV RC by binding to WIPI2B and CAPRIN1, and that IFN gamma-mediated control of MNV replication was dependent on CAPRIN1. Collectively, this study identifies a novel mechanism for the autophagy machinery-mediated recognition and inhibition of viral RCs, a hallmark of (+) RNA virus replication.

ATG7, and an E2-like conjugating enzyme, ATG3, function together with an E3-like ligase, the ATG12-ATG5-ATG16L1 complex, to conjugate LC3 to PE (7). Intriguingly, the LC3 conjugation machinery is required for the inhibition of the (1) RNA virus MNV by interferon gamma (IFNG), which specifically targets the replication stage of viral life cycle (8,9). LC3 accumulates at viral RCs in MNV-infected cells in an ATG5 and ATG16L1-dependent manner (8,9); then, IFN-inducible GTPases, including immunityrelated GTPases (IRGs) and GBPs, target the LC3-marked RCs upon their induction by IFNG, resulting in the inhibition of viral replication (8,9). We found the same role of autophagy machinery in controlling parasitophorous vacuole membrane of Toxoplasma gondii (10) and termed this process Targeting by AutophaGy proteins (TAG) (11). However, it remains unknown how the ATG12-ATG5-ATG16L1 complex detects RC and recruit GTPases to the replication site.
IFNG is a multifunctional cytokine that regulates a broad spectrum of immune networks (12). In contrast to the type I and III interferons produced by any cells upon viral infection, IFNG expression is restricted to specific subsets of immune cells upon infection or stimulation with antigens, but almost all cells can respond to IFNG (13). Although IFNG has now been shown to play diverse roles in immune modulation and defense against a broad-spectrum of pathogen, it was originally discovered, and named, as a soluble factor that interferes with viral replication (14). In addition to immune modulating function, which mediates broad immune defense against infection, IFNG can play a role in direct antiviral effects on infected cells and neighboring cells (15). By inducing multiple downstream interferon-inducible genes and targeting specific viral components, IFNG has shown direct antiviral activity and can target different stage of virus life cycle to inhibit virus infection. Despite the observations about IFNG direct antiviral activity, the detailed mechanisms of IFNG have not been well characterized.
In this study, we investigated two aspects of the TAG system: (i) Do the human homologs of the mouse GBPs also have antiviral activity via TAG and what functional domains are required for TAG? (ii) How does the LC3 conjugation machinery recognize MNV RCs? We observed a conserved role for mGBP2/hGBP1 in TAG antiviral activity. We also discovered a role for WIPI2 recruitment of ATG16L1 to RCs and uncovered a novel role for CAPRIN1 in recruiting ATG16L1 to RCs.

RESULTS
Human GBP1 localizes to the MNV replication complex and is sufficient for IFNG-mediated control of MNV. We previously showed that human GBPs target MNV RCs upon their formation in human cells (8). There are 23 IRGs, 11 GBPs, and two pseudogenes in the mouse, and mouse IRGs play a crucial role in GBP function. In contrast, only 2 IRGs and 7 GBPs have been identified in humans, in which hIRGs are not induced by IFNG (16). To understand the antiviral mechanism of human GBPs, we investigated which hGBP plays a crucial role against MNV.
We examined the localization of all seven human GBPs individually, with or without IFNG stimulation. We used HeLa cell line expressing individual HA-tagged GBPs or COVA (cytoplasmic ovalbumin) which was used as a negative control. We transfected these cells with a plasmid expressing MNV ORF1 for 6 h followed by treatment with or without IFNG at 100 U/mL for an additional 18 h and tested colocalization of HA-GBPs with the MNV RC. Only HA-GBP1 showed significant colocalization with MNV RC in the presence of IFNG, suggesting that an IFNG-induced cofactor is necessary for targeting GBP1 to the viral RC ( Fig. 1A to C). There was various expression of HA-GBP proteins in transduced HeLa cells, with GBP7 having the lowest expression compared to other GBPs (Fig. 1B). Although the differential GBP expression should be considered when interpreting the experiment, HA-GBP1 clearly colocalizes with MNV RC to significantly higher levels than other GBPs. We also observed a significant reduction of IFNG control of MNV replication in GBP1 2/2 and ATG16L1 2/2 HeLa cells compared to WT cells ( Fig. 1D and E), demonstrating that ATG16L1 (as already shown in [8]) and GBP1 function in IFNG control of MNV replication in human cells.
GBPs belong to the dynamin family, with nucleotide-dependent oligomerization and GTPase activity, and has a C-terminal isoprenylation anchor that enables membrane binding. To determine which GBP1 functions are responsible for targeting and inhibiting MNV RC, we generated various GBP1 mutants, including GTP hydrolysis-defective (GBP1/R48A), GTP binding-deficient (GBP1/D184N), and membrane association impaired (GBP1/C589A) (17,18). While all GBP1 mutants were expressed at comparable levels, they failed to target the MNV RC formed by transfecting MNV ORF1 into HeLa cells ( Fig. 2A and B). Knockout (KO) of GBP1 interfered with the antiviral function of IFNG, and IFNG control of MNV replication was restored when GBP1 expression was reconstituted. However, reconstitution of GBP1 expression with either the GBP1 mutants or the negative control failed to completely FIG 1 Human GBP1 localizes with the MNV replication complex and is important for IFNG-mediated control of MNV replication. (A) Immunofluorescence assay (IFA) for the localization of GBPs to the MNV RC in HeLa cells expressing individual GBPs. The cells were transfected with a plasmid expressing MNV ORF1 for 6 h followed by treatment without or with IFNG at 100 U/mL for an additional 18 h. Quantitation of IFA for the localization of GBPs (1-7) on to the MNV RC is shown (B) Representative Western blot of HeLa cells from (A); expressing COVA (cytoplasmic ovalbumin; used as a control) and individual GBPs. (C) IFA for the localization of GBPs to the MNV RC. HeLa cells expressing individual GBPs were transfected with a plasmid expressing MNV ORF1 for 6 h followed by treatment without or with IFNG at 100 U/mL for an additional 18 h. Representative images shown here. Zoomed images shown only for cells showing colocalization. Scale bar, 5 mm. (D) As human cells do not express MNV receptor, MNV vRNA were transfected into the WT, ATG16L1 2/2 and GBP1 2/2 HeLa cells and MNV production was measured by TCID50. (E) Representative Western blot of HeLa cells for (D); WT, ATG16L1 2/2 and GBP1 2/2 . Cells were untreated or treated with 100 U/mL IFNG for 24 h. Data were analyzed using unpaired t test. ND, not detected; ns, not significant; *, P , 0.05. Dashed lines in D indicate the limit of detection (LD).
Host Detection of Viral Replication Complexes mBio restore IFNG antiviral activity (Fig. 2C). These results suggest that the known intrinsic properties of GBP1 are required for targeting to viral RC and inhibition of MNV replication by IFNG.
The GBPs have similarities with dynamin-like GTPases and undergo self-assembly to form large complexes. In the case of several GBPs, including GBP1, prenylation of a Cterminal CaaX motif enables an association to the membrane. The GBPs can also heterodimerize with other members of the GBP family resulting in hierarchical positioning on the intracellular vesicles and recruiting nonprenylated GBPs to their subcellular compartments (17,19,20). These properties raise the question of whether GBP1 alone plays a role in GBP1 function in IFNG-mediated control of MNV or whether other GBPs also participate. We previously observed that MNV replication was substantially less inhibited by IFNG in human HAP1 cells with ATG16L1 2/2 or GBP1-7 2/2 (whole GBP KO [21]) than in the control cells (8). GBP1 expression in ATG16L1 2/2 HAP1 cells cannot rescue IFNG-mediated control of MNV, as expected, because it lacks the LC3 conjugation system required for GBP targeting to the MNV RC ( Fig. 2D and E). In contrast, GBP1 FIG 2 GBP1 is the necessary and sufficient human GBP for IFNG inhibition of MNV replication. (A) Quantitation of IFA for the localization of GBP1 mutants on to the MNV RC. HeLa cells expressing GBP1 and mutants were transfected with a plasmid expressing MNV ORF1 for 6 h followed by treatment with IFNG at 100 U/mL for an additional 18 h. After incubation, the cells were fixed and stained with anti-Propol antibody and anti HA antibody for GBP1 and its mutants (B) Representative Western blot of HeLa cells to confirm the protein reconstitution for (A and C). (C) MNV production in the WT and GBP1 2/2 HeLa cells expressing either COVA (control) or WT/mutants GBP1. (D) Representative Western blot of WT, ATG16L1 2/2 and GBP1-7 2/2 HAP1 cells expressing either COVA (control) or GBP1 for (E). Cells were untreated or treated with 100 U/mL IFNG for 24 h (E) MNV production in the WT, ATG16L1 2/2 and GBP1-7 2/2 HAP1 cells expressing either COVA (control) or GBP1. For all MNV production measurement in HeLa cells, each group of cells were untreated or treated with the 100 U/mL of IFNG for 24 h and then transfected with 50 ng MNV viral RNA. At 24 hpt, cells were harvested to titer infectious virus by TCID 50 . Data were analyzed using unpaired t test. ND, not detected; ns, not significant; **, P , 0.01; ***, P , 0.001; ****, P , 0.0001. Dashed lines in C, D indicate the limit of detection (LD). ND, Not Detected.
Host Detection of Viral Replication Complexes mBio expression in GBP1-7 2/2 HAP1 cells restored the IFNG-mediated inhibition of MNV replication. Taken together, these results show that GBP1 is the necessary and sufficient human GBP for IFNG inhibition of MNV replication. ATG16L1 colocalizes with MNV and HNV NS4-derived RCs. Previously, we found that individual expression of ATG16L1, but not ATG5 or ATG12, is sufficient to localize to the MNV RC, suggesting that it interacts with a component of the RC and recruits the ATG12-ATG5-ATG16L1 complex to the RC (9). ATG16L1 is a well-known scaffold protein that primarily functions through protein-protein interactions (22). Therefore, we hypothesized that the interaction of ATG16L1 with other protein(s) would be required for the viral RC targeting of ATG16L1. We observed that ATG16L1 is localized to the RC formed upon MNV ORF1 expression (9) (Fig. 3A), consistent with our previous report that LC3 localizes with RC-like structures formed by MNV ORF1 expression (8).

Host Detection of Viral Replication Complexes mBio
We also observed that expression of either MNV or human Norwalk virus (HNV) NS4, a component of ORF1 that is sufficient to induce RC-like structures (23), can colocalize with ATG16L1 ( Fig. 3B and Fig. S1A). In addition, LC3 and GBP1-5 also colocalized with MNV NS4 puncta (Fig. S1B).
We generated truncation mutants of ATG16L1 to identify important domains for ATG16L1 targeting to viral RCs; full length (FL), DWD40 repeats (1-249), F32A/I35A/I36A that is defective for the direct membrane binding capacity (1-249-FII), DFIP200 binding (1-230), DWIPI2 binding (1-207), and the coiled-coil domain (79-249) (Fig. 3C). We transduced these constructs in Atg16l1 f/f (Atg16l1 WT) and Atg16l1 f/f 1 LysMcre (Atg16l1 KO in myeloid cell lineage) bone marrow-derived macrophage (BMDM) cells and measured the restoration of IFNG-mediated control of MNV. In Atg16l1 WT BMDMs, ATG16L1 expression did not affect the MNV control by IFNG, similar to the untransduced and EGFP-transduced negative controls ( Fig. 3D and S1C). In Atg16l1 KO BMDMs, MNV replication was not inhibited by IFNG, while ATG16L1 FL expression restored IFNG-mediated control of MNV replication ( Fig. 3E and S1C). We observed that the ATG5 binding region and WIPI2 binding region (207-230) of ATG16L1 were essential for IFNG-mediated control of MNV in Atg16l1 KO BMDMs ( Fig. 3C and E). We next evaluated the ability of the ATG16L1mutants to promote the localization of LC3 with the MNV RC in infected cells. As expected, these results mirrored the IFNG inhibition of MNV replication in that the DATG5 binding region (i.e., the coiled-coil domain) and DWIPI2 binding (1-207) mutant of ATG16L1 did not restore the LC3 lipidation on MNV RC ( Fig. 3F and Fig. S1D). Collectively, these data corroborate the importance of ATG5 binding region of ATG16L1 in LC3 lipidation on virus RC and MNV control by IFNG (8) and further demonstrate that the 208-230 region of ATG16L1 is required for LC3 conjugation to the viral RC and IFNG-mediated control of MNV.
WIPI2 contributes to LC3 localization with MNC RC. The 208-230 region of ATG16L1 contains a WIPI2 interaction domain, which leads to recruitment of ATG16L1 to the membrane surrounding bacteria and subsequent restriction of bacterial proliferation (25). To test whether WIPI2 localizes with LC3 and the viral RC, we transfected MNV ORF1 and WIPI2-RFP into HeLa cells and examined the WIPI2-LC3-MNV Propol colocalization. Under the cotransfected condition, both WIPI2-RFP and LC3 localized on MNV RCs with high Mander's coefficients (Fig. S2A). To test whether WIPI2 is required for LC3 colocalization with MNV RCs, WT and Wipi2 2/2 mouse embryonic fibroblasts (MEFs) expressing an MNV receptor CD300LF (29,30) were infected with MNV at MOI 10 for 10 h and fixed/stained with anti-Propol and anti-LC3 antibodies. Under these conditions, LC3 localization with MNV RC was observed in WT MEF; this localization was dependent on the presence of Wipi2, as their localizations were substantially reduced upon Wipi2 deletion (Fig. S2B to D). Notably, some LC3 localization with MNV Propol remained in Wipi2 2/2 MEFs, suggesting that WIPI2 was not absolutely required for LC3 colocalization with RC, and implicating a potential role of other ATG16L1-interacting proteins in RC targeting. Further, we examined the role of WIPI2 in IFNG-mediated inhibition of MNV replication. Here, we used BV2 cells, which can be readily infected by MNV through endogenous CD300LF. Intriguingly, MNV replication was controlled by IFNG in Wipi2 2/2 BV2 cells, similar to the WT control ( Fig. 4A and B). Thus, although Wipi2 deletion resulted in decreased colocalization of LC3 with viral RC, it did not result in impaired IFNG control of MNV infection in BV2 cells, suggesting the possibility of functional redundancy.
WIPI2 recruits the LC3 conjugation system to the autophagosome membranes through its association with phosphatidylinositol 3-phosphate (PI3P) enriched membranes (25). In addition, PI3P localized with the RCs of some RNA viruses, facilitating their replication (31,32). We established Cd300lf-MEF cells stably expressing 2xFYVE-EGFP, which binds PI3P. PI3P production was significantly inhibited by wortmannin (a Pan-PI3K inhibitor that blocks PI3P production) compared with control cells (Fig. S3A), but we did not find any difference in LC3 localization with MNV RC upon wortmannin treatment ( Fig. S3B and C). Further, we examined the role of IFNG in WT and VPS34depleted MEFs, which is a major kinase for generating PI3P (33), using Vps34 f/f MEFs (34). Consistent with the localization of LC3 independent of PI3P production, VPS34deficiency did not show any effect on IFNG-mediated virus regulation ( Fig. S3D and E). Collectively, these data show that WIPI2 is involved in LC3 localization onto the MNV RC but is not essential for IFNG-mediated MNV control.
The ATG16L1-interacting protein CAPRIN1 is important for IFNG-mediated control of MNV replication. Given that WIPI2 was not essential for IFNG inhibition of MNV replication, we tested a role for other ATG16L1-interacting proteins in MNV-infected cells. To minimize identifying TAG-unrelated interacting proteins, we used the 1-230 ATG16L1 truncation mutant, which was functional for LC3 conjugation on RC and IFNG-mediated MNV control (Fig. 3E). We performed immunoprecipitation-mass spectrometry (IP-MS) (35) and identified proteins interacting with 1-230 ATG16L1 in uninfected and MNVinfected BV2 cells. We listed proteins found in both samples and highlighted the unique proteins that showed significantly higher signal from the MNV infected sample (Table S1).
It was recently reported that CAPRIN1 interacts with MNV RC during infection in replication-independent manner via its interacting partner, G3BP1 (49). Based on our IP-MS data with ATG16L1, we hypothesized that ATG16L1 complex might be recruited to the virus RC via interaction with CAPRIN1. We confirmed that ATG16L1 interacts with CAPRIN1 via co-IP (Fig. 5F) and further observed that CAPRIN1 colocalized with MNV RC and ATG16L1 (Fig. 5G). These data suggested that CAPRIN1 can bind to ATG16L1 and colocalize with MNV RC upon infection. Further, we found that LC3 localization with MNV RCs was dependent on CAPRIN1 expression and reduced LC3 localization on MNV RC by CAPRIN1 KD was successfully restored by CAPRIN1 reconstitution ( Fig. 5H and I). Collectively, these results suggest that CAPRIN1 is required for LC3 localization with the MNV RC and IFNG-mediated control of MNV replication.

DISCUSSION
In this study, we found that human hGBP1 is sufficient to localize with MNV RCs and to inhibit MNV replication upon IFNG stimulation. This is supportive of our previous finding that murine mGBP2, the homolog hGBP1, is required for IFNG control of MNV infection in the mouse system. The inability of hGBP1/mGBP2 to localize with MNV RCs in the absence of IFNG stimulation suggests that additional IFNG-induced factors are involved in recruiting GBPs to MNV RCs. The ability of GBP1 to inhibit MNV replication depends on its core properties of GTP binding, hydrolysis, and membrane binding. This suggests that dynamin-like activity of GBP, which multimerizes on the endomembranes and a pinches off small vesicles (50), plays a role in the control of RC via TAG. It should be noted that the relative contribution of the TAG system in IFNGmediated control of MNV was more pronounced in murine cells than in human cells (e.g., Fig. 3E and 1D, respectively). This may reflect that human cells have an evolutionarily contracted IRG system (8,51). How exactly GBP1 disrupts the MNV RC is the question of future investigation.
In autophagy, ATG16L1 determines the site of LC3 lipidation on the developing phagophore. The ATG16L1 protein, which has three main domains (N-terminus, coiled coil, and WD40 repeat), can drive LC3 lipidation on double-membrane autophagosomes and single membrane vesicles. We found that the ATG5-binding region of ATG16L1 was essential for LC3 lipidation of viral RC, which was expected, as the ATG12-ATG5-ATG16L1 complex as a whole is responsible for bringing LC3 to the developing phagophore (52). ATG16L1 is recruited to the developing phagophore by interacting with newly synthesized PI3P on the phagophore. This recruitment is mediated by either direct interaction of ATG16L1 with phosphoinositide or by binding to the PI3P effectors WIPI2 or FIP200 (24)(25)(26)(27). We found that ATG16L1 region 207-230, which is known to bind WIPI2, is critical for recruiting ATG16L1 to the MNV RC, potentially implicating WIPI2 in RC recognition. We observed that WIPI2 localized to the MNV RC and affected LC3 localization with RCs, but intriguingly it was not required for IFNGmediated control of MNV in BV2 cells. It is possible that the reduced level of LC3 localization with RCs in WIPI2 2/2 cells is still sufficient for IFNG control of MNV replication. Alternatively, WIPI2 function may be secondary to CAPRIN1 in IFNG control of MNV replication. Nevertheless, another study (53) suggested a crucial role of WIPI2 in the control of MNV replication by IFNG. This discrepancy might stem from differences in experimental systems. Unlike canonical autophagy, the PI3K activity, which promotes PI3P production, is not required for TAG-mediated IFNG control (9, 10); consistently, we found that VPS34 activity and PI3P were not required for TAG-dependent MNV control, demonstrating that this regulation is independent of PI3P. Overall, our data suggest that WIPI2 plays a crucial role in LC3 localization on MNV RC but is not essential for MNV inhibition by IFNG.
Our previous studies showed that LC3s are necessary to recruit the IFN-inducible GTPases to the RCs (8). Therefore, we speculated there might be a host protein that recruits ATG16L1 to the membrane of RCs and is required for IFNG inhibition of MNV replication. The conditions we considered to find a promising protein candidate were 1) a protein that interacts with ATG16L1 and 2) a protein that has the potential to interact with the replication compartment membrane at the same time. To explore the candidate proteins that meet these criteria, we searched the interactome of the functional ATG16L1 1-230 in MNV-infected cells. We found that CAPRIN1 is an ATG16L1-interacting protein required for TAG-dependent IFNG control of MNV infection. CAPRIN1, also known as RNA granule protein 105 (54), is an RNA-binding protein (RBP) and participates in stress granule (SG) formation through phosphorylation of eIF2a (55). CAPRIN1 regulates protein translation, cell migration, and proliferation in various cell types through interactions with several SG-associated RBPs, such as G3BPs and/or USP10 (55)(56)(57). Viruses interfere with host defense mechanisms against viral replication through binding with various RBPs, including CAPRIN1 (37,(58)(59)(60)(61). During MNV infection, it was confirmed via IP that ATG16L1 interacts with CAPRIN1, which is presumed to be mediated by G3BP1, an interacting partner of CAPRIN1 (49,62). Of note, CAPRIN1 interacts with ATG16L1 and G3BP1 in uninfected cells, and they are relocalized to MNV RC upon infection (49). We found that CAPRIN1 not only interacts with ATG16L1 but also is required for the IFN-mediated control of MNV replication.
The ATG16L1 1-230 interactome included several other RBPs in addition to CAPRIN-1. RNA viruses coopt and directly interact with host RBPs to affect the recruitment of viral genomes for replication, to help the membrane assembly for the replication compartment, to control RNA synthesis, and to make viral RNA stable (63). Since viral RC tagging by ATG16L1 is not limited to MNV, whether these RBPs, including CAPRIN-1, generally affect the TAG system is unclear. Also, CAPRIN1 involvement in the control of other (1) RNA virus RC, like encephalomyocarditis virus (8), needs further investigation.
In summary, our study demonstrates that the ATG16L1 is a critical molecule to target the RC of the MNV and IFNG-mediated antiviral effect as a member of the LC3-conjugation system. This system was conserved in humans, targeting both MNV RC and HNV RC, and hGBP1 was sufficient to inhibit MNV replication by IFNG. Importantly, we identified CAPRIN1 as a previously unappreciated ATG16L1-interacting protein involved in recruiting the LC3 lipidation complex to RCs and required for the inhibition of MNV replication by IFNG (Fig. 6). This study warrants further investigations to understand the cell autonomous immune defense mechanism against viral RCs, the hallmark of all known (1) RNA viruses.

MATERIALS AND METHODS
Mice. Atg16l1 f/f 1LysMcre mice is previously described (8). In all experiments, control and experimental mice were 6 to 8 week old littermates of both genders. Littermates of the same sex were housed together regardless of genotype. All mice were housed and bred at the University of Chicago under specific-pathogen-free conditions in a biosafety level 2 facility in accordance with federal and university guidelines. All experimental procedures were approved by the Animal Care and Use Committee of the University of Chicago.
Primary cells. The primary cells utilized in this study were murine bone marrow derived macrophages (BMDMs). Bone marrows were isolated from the femurs and tibias of 6 to 8 week old mice of both genders listed above and plated in nontissue culture treated 10-cm dishes in 10 mL of BMDM media. On day 4, 10 mL of fresh BMDM medium was added. On day 7, BMDMs were detached from the dish using ice-cold 0.02% EDTA in DPBS (Sigma-Aldrich, E8008), and seeded in tissue culture treated plates or on coverslips for subsequent experiments. Remaining D7 BMDMs were frozen in BMDM media containing 10%DMSO and used later to set up replicate experiments if necessary. BMDMs were rested for 3 days after seeding and used for experiments. The composition of BMDM medium was Dulbeco's Modified Eagle Medium (Mediatech, 10-013), 10% fetal bovine serum (Biowest, US1520), 5% horse serum (Life Technology,  and 10% fetal bovine serum (Biowest, US1520) at 370C under 5% CO2. All transfections were done using Lipofectamine 3000 (Thermo Fischer-L3000001) as per the manufacturer's instructions. CRISPR KOs of Wipi2 and GBP1. The gene-specific guide RNA was cloned into lentiCRISPRv2 vector and cotransduced into BV2 cells with the packaging plasmids pVSVg and psPAX2. Forty-eight hours after transfection the lentiviral particles in the cell culture media were applied to cells for 48 h. The transduced cells were then selected with puromycin at 3 mg/mL for 48 to 72 h. Cells were seeded in single cell density and successful KO clones were confirmed by immunoblotting using antibody specific for Wipi2. For GBP1 KO in HeLa cells, gRNA was cloned in px458 (number 48138) and px459 (number 48139) plasmids that were obtained from Addgene. Cells were transfected with the plasmids using Lipofectamine 3000. After 48 h, transfected cells were enriched through one passage of puromycin (Sigma-Aldrich, P9620) treatment (3 mg/mL). Enriched cells were diluted and seeded in 96-well plates, aiming for 0.5 cell/well, and the monoclonal colonies of cells were picked at 7 days after the set-up and further amplified. The guide RNAs are listed in Table S2.
Viral infection and transfection. All MNV infections were conducted using the MNV-1.CW3 strain (8). MNV-1.CW3 viruses were prepared from a cDNA clone containing the genome of MNV-1.CW3. 1 Â 10 6 293T cells were seeded in 6-well plates and transfected with 4 mg of the MNV-1.CW3 plasmid for 48 h to produce virus. MNV was further amplified in BV-2 cells. Infected cells were incubated until the cells showed . 90% cytopathic effect, usually for 48 h. Infected cells were then frozen and thawed, and the cell lysates containing viral particles were centrifuged for 20 min at 3000 rpm to remove the cell debris. Supernatants were further centrifuged for 3 h at 26,250 rpm at 4°C to produce a concentrated virus stock. The viral stocks were frozen in small aliquots, and the titers of the stocks were determined by TCID 50 method as described earlier (8). All viral infections were conducted with a viral stock aliquot having undergone only one cycle of freeze and thaw. MNV vRNA was isolated from the concentrated MNV stocks using TRI reagent (Sigma-Aldrich, T9424) according to the manufacturer's instruction. As human cells do not express MNV receptor, infection studies were conducted by transfecting MNV vRNA. The amount of vRNA indicated in figure legends was transfected to cells using Lipofectamine 3000 (Thermo Fischer-L3000001) as per manufacturer instruction manual.
Lentiviral transduction. All lentiviruses were generated in 293T cells as previously described (8). The lentiviral vector plasmids (psPAX2; Addgene, 12260) and pseudotyping vector (pMD2.G; Addgene, 12259) along with gene construct were transfected in 293T cells using the calcium phosphate precipitation method. Produced lentivirus was filtered through a 0.45 -mm syringe filter (Millipore, MA) and added onto the cells.
TCID 50 assay. TCID 50 assay was performed to quantitate infectious MNV as described previously (9). Serial 10-fold dilution of viral lysates were added to BV2 cells seeded in 96-well format. 8-wells were infected for each dilution and further incubated for 5 days. TCID 50 was calculated by determining the dilution factor needed to show cytopathic effect (CPE) in 4 out of 8 (50%) wells. The limit of detection was calculated as the amount of virus that cause CPE in 4 out of 8 wells at the lowest dilution. Western blot. Cells were harvested in sample buffer (0.1 M Tris [pH 6.8], 4% SDS, 4 mM EDTA, 286 mM 2-mercaptoethanol, 3.2 M glycerol, 0.05% bromophenol blue), and proteins were analyzed as previously described (8). Commercial antibodies used in this study are as follows: MNV Propol (9) ATG16L1 and LC3B (Sigma-Aldrich); GBP1-5 and Actin-HRP (Santa Cruz Biotechnology); HA (the Frank W. Fitch Monoclonal Antibody Facility, The University of Chicago); Flag-M2 (Thermo Fischer); WIPI2 (Abcam); HRP Goat anti-mouse and HRP Donkey anti-rabbit (Bio Legend); and HRP Donkey anti-goat (Jackson Immuno Research).
Immunofluorescence's analysis. For immunofluorescence analysis cells were grown on cover glass (Fisher Scientific, PA; 12-545-80) in 24-well plates and then untreated/treated/infected as described in figure legends. Cells were fixed in 4% formaldehyde (Ted Pella, 18505), cells were permeabilized with 0.05% Saponin (Acros, 41923) in PBS, and blocked and probed in PBS containing 0.05% Saponin and 5% normal donkey serum (Jackson Immuno Research, 017-000-121). Fixation, permeabilization and 2 rounds of blocking were conducted for 10 min each at room temperature and staining with primary and secondary antibodies was conducted for 1 h each at room temperature. Samples were washed 5 times with PBS/0.01% Saponin for 5 min after each antibody incubation. Nuclei were stained with Hoechst 33342 (Invitrogen, H1399). Coverslips were mounted on glass slides with ProLong Diamond Antifade Mountant (Invitrogen, P36961). The images were acquired using Olympus DSU confocal microscope with a 60Â water objective. Digital images were taken with Slidebook 6.0 software and images from each channel were merged using ImageJ (National Institutes of Health). Analysis of images was performed with ImageJ and used a set of defined intensity thresholds on all images. Primary antibodies detecting the following proteins were used: MNV Propol (9) qRT-PCR. Total RNA was harvested from cells using RNeasy 96 kit (Qiagen) for measuring KD efficiency upon shRNA transduction. Cellular RNAs were reverse transcribed, and PCR amplified using the SuperScript III Platinum One-Step qRT-PCR System with Platinum Taq (Invitrogen) and IDT Primer Assays (Integrated DNA Technologies). Cellular RNAs were normalized to 18S levels using StepOnePlus System (Applied Biosystems).
Mass spectrometry. BV2 cell line stably expressing HA tagged 1-230 ATG16L1 truncation mutant was generated by lentiviral transduction. These cells were mock treated or infected with 10 MOI of MNV for 10 h. Cell lysate was pulled down on HA beads and sent to the Taplin Biological Mass Spectrometry Facility at Harvard Medical School for further analysis of interacting partners by mass spectrometry.
Quantification and statistical analysis. The mean with SD were calculated and described for all TCID50 and qRT-PCR data using two or more biological replicates. The Standard Error of Mean (SEM) were calculated and described for all IFA quantification and Mander's coefficients calculation. All data were analyzed with Prism software (GraphPad) using one-way or two-way analysis of variation (One-way or two-way ANOVA) with multiple comparisons (for multiple samples), unpaired t test (for two samples) as indicated in figure legends. No specific method was used to determine whether the data met assumptions of the statistical approach. All differences not specifically indicated as significant were not significant (n.s., P . 0.05). Significant value was indicated as *, P , 0.05; **, P , 0.01; ***, P , 0.001; ****, P , 0.0001. Details of statistical significance and n values can be found in the figures or corresponding figure legends.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.   We have no conflicts of interest to declare.