The autophagy machinery interacts with EBV capsids during viral envelope release

Significance

Viruses have evolved mechanisms to exploit autophagy for their benefit. Epstein–Barr virus (EBV) was found to carry LC3B-II, a hallmark of autophagy, in its virions, suggesting recruitment of autophagic membranes to its envelope. However, the viral proteins mediating the interaction with LC3B have remained elusive. Here, we analyzed the virus proteome and found numerous viral and autophagy proteins present in extracellular virions, confirming a role for autophagy during virus egress. Moreover, we identified additional functions for capsid scaffold proteins BVRF2 and BdRF1 by interacting with LC3B-membranes and facilitating the release of viral particles. This study highlights a role for EBV proteins interacting with autophagic membranes and requirement for the LC3B conjugation complex of ATG5-ATG12-ATG15L1 for viral envelope release.

Abstract

Autophagy serves as a defense mechanism against intracellular pathogens, but several microorganisms exploit it for their own benefit. Accordingly, certain herpesviruses include autophagic membranes into their infectious virus particles. In this study, we analyzed the composition of purified virions of the Epstein–Barr virus (EBV), a common oncogenic γ-herpesvirus. In these, we found several components of the autophagy machinery, including membrane-associated LC3B-II, and numerous viral proteins, such as the capsid assembly proteins BVRF2 and BdRF1. Additionally, we showed that BVRF2 and BdRF1 interact with LC3B-II via their common protein domain. Using an EBV mutant, we identified BVRF2 as essential to assemble mature capsids and produce infectious EBV. However, BdRF1 was sufficient for the release of noninfectious viral envelopes as long as autophagy was not compromised. These data suggest that BVRF2 and BdRF1 are not only important for capsid assembly but together with the LC3B conjugation complex of ATG5-ATG12-ATG15L1 are also critical for EBV envelope release.

Macroautophagy (hereafter called autophagy) is a ubiquitous catabolic pathway characterized by the engulfment of cytoplasmic content into a double-membrane vesicle, called autophagosome, which ultimately fuses with lysosomes for degradation (1, 2). Several AuTophaGy-related (ATGs) genes and their encoded proteins coordinate this catabolic pathway, including the mammalian microtubule–associated protein 1 light chain 3 (LC3). This protein family of six members in humans (MAP1LC3A, MAP1LC3B, MAP1LC3C, GABARAP, GABARAPL1, and GABARPL2) labels autophagosomes (3). The formation of autophagosomes requires two ubiquitin-like conjugation systems that will conjugate LC3 to their membranes, supporting their elongation and autophagy substrate recruitment. In the first ubiquitin-like system, the proform of LC3 is cleaved by ATG4, generating cytosolic LC3-I. LC3-I is subsequently conjugated to phosphatidylethanolamine in the membrane of growing autophagosomes by ATG7, ATG3, and ATG12-ATG5-ATG16L1, the latter of which is generated by the second ubiquitin-like system of ATG7 and ATG10 conjugating ATG12 to ATG5 (4). The membranes that will be decorated with LC3 are first labeled with phosphoinositol-3-phosphate deposited by the PI3 kinase VPS34 in a complex with BECN1 and ATG14. This complex in turn is activated by the protein kinase ULK1 in complex with ATG13, FIP200, and ATG101, which is activated by nutrient starvation. The lipidated form of LC3 is referred to as LC3-II and is present at both external and luminal membranes of the autophagosome. Upon its fusion with the lysosome, the sequestered content as well as the inner membrane with LC3-II is degraded by lysosomal hydrolases (3). Furthermore, autophagy can selectively target intracellular pathogens for degradation, a process termed xenophagy, Autophagy receptors, such as sequestosome1 (SQSTM1/P62) or neighbor of BRCA1 gene 1 (NBR1), recognize and bind eat-me signals, like ubiquitin or galectins, decorating pathogens or ruptured endosomes, while simultaneously interacting with LC3 and therefore deliver pathogens to autophagic degradation (5). Thus, xenophagy serves as a cell-intrinsic defense mechanism against intracellular pathogens. However, microorganisms have coevolved with autophagy to subvert and escape this degradative pathway or even exploit it for their own benefit and survival, such as in the case of herpesvirus infections.

Epstein–Barr virus (EBV) is a ɣ-herpesvirus that infects more than 95% of the adult human population (6). EBV was the first human tumor virus to be discovered. Its two-step life cycle comprises a latent stage for viral replication by host cell proliferation and for viral persistence, and a lytic cycle to produce infectious virions (7). The process of capsid formation is at the core of infectious virus production. It begins for EBV by the formation of a capsid scaffold in the nucleus, mainly driven by the two proteins BVRF2 and BdRF1 (8, 9). These proteins together with the major capsid protein (MCP), also known as viral capsid antigen (VCA), coordinate the addition of other capsid proteins, such as BDLF1, BORF1, and BFRF3 (P18) (10, 11). Mature capsids cross the inner and outer nuclear membranes to reach the cytoplasm, where they acquire a tegument protein layer and are enclosed in membranes of cellular origin. These membranes are decorated with glycoproteins, such as GP350 and form the viral envelope (12, 13). Indeed, the trans-Golgi network, endosomal compartments or autophagosomes have been proposed as potential membrane donors for the EBV envelope, enclosing tegumented capsids into double-membrane vesicles that travel to the plasma membrane (14–16). The outer membrane fuses with the cellular membrane, releasing the mature virion to the extracellular space. We and others have demonstrated that EBV manipulates autophagy during its lytic stage and that autophagy is important for virus secretion (17–20). Autophagy-deficient cells show an impairment of virus production, while accumulating viral DNA in the cytoplasm, suggesting a role for autophagy in envelope acquisition. We previously showed that the marker of autophagosomes, LC3B-II, is associated with EBV particles, suggesting autophagic membranes as one of the sources of EBV envelopes (17). However, the molecular mechanism driving the interaction between EBV capsids and the autophagy machinery has not been explored so far, nor for any other virus that utilizes LC3 decorated membranes for its envelope (21).

Here, we identify the EBV capsid assembly proteins BVRF2 and BdRF1 together with the LC3B lipidation machinery in infectious viral particles. We describe that they bind LC3B-II via their common C-terminal region and are required in addition to LC3B lipidation for the egress of viral envelopes. We propose that this additional BVRF2/BdRF1 function supports efficient virion production.

Results

Viral and Autophagy Proteins Are Contained in EBV Virions.

Herpesviruses are believed to acquire their viral envelopes from perinuclear vesicular compartments (22). Indeed, we have previously shown that the lipidated form of LC3B (LC3B-II), a hallmark of autophagosomes, was found associated with extracellular EBV virions (17). However, whether other proteins associated with the autophagy machinery are also incorporated into EBV virions has not yet been addressed. To further assess the protein content of EBV virions, EBV-producing cells HEK293/EBV(WT) [referred to as EBV(WT)] were induced to the lytic cycle (Lytic) or left uninduced (UI). Extracellular vesicles (EV), enriched for EBV virions, were concentrated by pelleting, followed by density gradient ultracentrifugation (Fig. 1A). The visible gradient-derived band contains high titers of infectious EBV in the lytic-induced condition, while few particles were detected in other fractions of the gradient (Fig. 1B and SI Appendix, Fig. S1A). Collecting smaller fractions of the density gradient revealed that fractions 2 and 3 also showed presence of EBV tegument protein BNRF1 (Fig. 1C and SI Appendix, Fig. S1B). In contrast, the absence of the endoplasmic reticulum markers Calnexin (Fig. 1C), P4HB as well as Grp94 (SI Appendix, Fig. S1 B and C) indicated that no contamination of cytoplasmic cellular proteins was found in the EV fractions enriched for infectious EBV. The autophagosome marker LC3B-II was also detected in EBV-enriched EV fractions after lytic cycle induction, which was absent in the UI control, consistent with our previous finding on the association of the membrane-bound LC3B-II with EBV particles (17).

Fig. 1.

Cellular and viral proteins identified in EBV virions. (A) Scheme of EV purification for MS analysis: Supernatants of EBV(WT) cells UI or induced to the lytic cycle were pelleted and further purified by density gradient ultracentrifugation. Bands containing EV enriched in infectious EBV particles were collected and either submitted to WB or dissolved in urea and digested in solution with Lys-C and trypsin endoproteases ON. Resulting peptides were purified by STAGE tips and analyzed by LC-MS/MS using data-independent acquisition (DIA). (B) Infectious EBV particles in visible EV band enriched for infectious EBV particles (band) and other fraction of the purification gradient (no band) produced in UI or lytic-induced cells. Expressed in green Raji units (GRU) per mL. Bars represent three independent experiments. (C) Whole-cell lysates (WCL) and the corresponding enriched EBV particles (purified EV band) were submitted to SDS-PAGE and immunostained for BNRF1, calnexin, and LC3B. Representative image from three independent experiments. (D) Principle component analysis based on the proteomic signature of all samples and the resulting variance between samples. Each datapoint represents the quantification of 3,565 proteins. (E) EBV proteins enriched in lytic samples versus control samples and categorized by structural location within the virus. (F) Hierarchical clustering showing DIA-based protein intensities, which were log2 transformed and z scored. Rows and column trees were hierarchically clustered using Euclidean distance. Two row clusters were identified in which protein abundances differed profoundly between EBV and control samples (clusters I and II). Significantly enriched GO terms of proteins in each cluster are annotated (FDR are listed in brackets). (G) Protein–protein interaction network based on STRING DB highlighting proteins related to autophagy, which were identified as significantly enriched in samples of enriched EBV particles (FDR < 0.05). Thickness of edges indicates strength of data support. Color of node indicates strength of enrichment, blue highlighting strongly enriched proteins. (H) WCL and the corresponding purified EV with enriched infectious EBV particles were submitted to SDS-PAGE and immunostained for BNRF1, BECN1, ATG5, and ATG7. Representative image from three independent experiments. (I) Scheme of the attachment assay. ELIJAH B cells were incubated with purified EV, enriched for infectious EBV particles, for 3 h at 4 °C to allow attachment of viral glycoprotein GP350 to the complement receptor 2 (CD21) on B cells. These cells are then placed on glass slides and stained for viral and autophagy proteins. (J) Purified EV from UI or Lytic cells were incubated with ELIJAH cells for attachment assay, followed by staining for DAPI (blue), P18 (green), VCA (red) and ATG5 (cyan). All confocal images are representative images from three different independent experiments. (Scale bar is 5 µm.) Quantitative analysis of VCA dots colocalizing with P18, ATG5 or both P18 and ATG5 in the attachment assays. Graph bar represents mean ± SEM of three independent experiments.

Next, EV enriched for infectious EBV particles were subjected to mass spectrometry (MS) analysis (Fig. 1 A and D–G) to identify viral and cellular proteins enriched in EV during lytic replication. Lytic and UI samples had distinct proteomic signatures (Fig. 1D), suggesting that the induction of the lytic cycle triggers substantial differences in the protein composition of EV, which include the presence of numerous EBV proteins (Fig. 1E). Fifty-two EBV proteins were enriched in lytic samples, including BcLF1 (MCP), BDLF1, BORF1, and capsid assembly proteins BVRF2/BdRF1; five of the six proteins required to self-assemble EBV capsids (23) (Fig. 1E). The EBV derived peptides detected by MS included the common region of BVRF2/BdRF1 that are encoded by the same viral open reading frame (ORF). All six capsid proteins were described to be present in EBV virions by a previously published mass spectrometry analysis (24) (Fig. 1E). However, proteins involved in processing and package viral DNA in the capsid were also detected in our analysis, such as BGLF1, BBRF1, BALF3, BDRF1, and BFLF1 (Fig. 1E). Similarly, tegument proteins like BLRFL2, BNRF1, BGLF4, and BRRF2, as well as glycoproteins BLRF1, BKRF2, BILF2, and BALF4, among others, are common hits between our and the previous analysis (24). Our analysis identified two additional tegument proteins, BBRF2 and BTRF1. Furthermore, we detected the presence of EBV proteins previously described to be associated with the structure of EBV but not involved in the formation of EBV virions, such as BMRF1 and BHLF1, but we also observed nonstructural proteins associated with viral EV, such as BZLF1, BNLF2b, and BKRF3 (Fig. 1E). For cellular proteins, hierarchical clustering of the enriched proteins allowed the identification of the gene ontology (GO) terms specifically enriched in each sample (Fig. 1F). Proteins belonging to GO terms viral reproduction, vesicular and endosomal transport, often mediated by Rab GTPases, and finally autophagy proteins were specifically enriched in EV during EBV replication. GO analyses by the term autophagy revealed several autophagy proteins enriched in EV with enriched EBV virions, including the already described LC3B protein (Fig. 1F). Additionally, key molecular players of autophagy such as members of the PI3 kinase complex (BECN1 and ATG14), the conjugation complex (ATG3, ATG7, ATG5, ATG4C, GABARAPL2) as well as autophagy receptors (P62 and NBR1), were all enriched compared to the control condition (Fig. 1G). The association of two of these hits with EV enriched in infectious EBV particles was confirmed by western blot, in which specific bands for BECN1, ATG5, and ATG7 were observed exclusively in EBV enriched EV (Fig. 1H). To confirm that these autophagy proteins are indeed part of EBV virions, and not associated with other vesicles of similar density, we established an immunofluorescence-based method to analyze single EBV particles. In this attachment assay, EBV enriched EV are incubated with ELIJAH cells to facilitate the attachment of EBV virions via the interaction of glycoprotein GP350 with complement receptor 2 (CD21) on B cells (Fig. 1I). ELIJAH cells showed a punctuate staining for capsid proteins VCA and P18 on their cellular membrane, and these two proteins colocalized to almost 100%, demonstrating that this technique can successfully stain viral particles that carry GP350 (to mediate attachment) via staining with either one of these capsid proteins (Fig. 1J). Additionally, VCA dots colocalized with ATG5 in approximately 64% of the events and had a 56% colocalization with both P18 and ATG5 (Fig. 1J), confirming the presence of the autophagy machinery in a subset of EBV virions. Additionally, we observed that EVs produced in ATG5-deficient cells, showed reduced P18-ATG5 colocalization (SI Appendix, Fig. S1F). Capsid proteins were also observed specifically colocalizing with LC3B and to a lesser extent with GABARAPL2, in comparison with the negative control RAB27A (SI Appendix, Fig. S1 D and E). These results suggest that EBV virions contain the LC3 family members LC3B and GABARAPL2 plus their lipidation complex.

EBV Lytic Proteins BVRF2 and BdRF1 Bind to LC3B.

EBV manipulates the autophagy pathway to promote the secretion of newly formed virions (17–20). However, how EBV interacts with autophagy at the molecular level remains to be explored. Therefore, we investigated which EBV proteins might interact with the autophagosome marker LC3B and mediate the incorporation of autophagy proteins into the virus particle. To identify LC3B-binding partners among EBV proteins, we established a proximity-dependent biotin identification (Bio-ID) technique in EBV(WT) cells, by stably expressing Myc-BirA*-LC3B or its control counterpart Myc-BirA*. The Bio-ID technique employs a promiscuous biotin ligase (BirA*), which biotinylates proximal endogenous proteins, allowing their selective tagging and subsequent isolation (25, 26) (Fig. 2A). We first confirmed the expression of the two constructs Myc-BirA*-LC3B and Myc-BirA* in EBV(WT) cells (Fig. 2B). Next, we verified the capacity of the transfected cells to produce infectious EBV at similar titers compared to unmodified cells (Fig. 2C). Furthermore, the ability of Myc-BirA*-LC3B to be lipidated and associated to autophagosomal membranes was validated under chloroquine treatment and upon induction of the lytic cycle, both described to cause autophagosome accumulation (3, 17) (SI Appendix, Fig. S2 A and B). Cellular and viral proteins that had an interaction with Myc-BirA*LC3B were successfully isolated by streptavidin affinity purification (AP), shown by the presence of streptavidin smear in the AP-Strep Myc-BirA* and Myc-BirA*-LC3B and by the absence of streptavidin signal in the supernatant (Fig. 2D). The isolated biotinylated proteins were analyzed by MS (Fig. 2E). LC3B and two well-known LC3B-binding proteins, ATG3 and FYCO1 (27, 28), were specifically enriched in Myc-BirA*-LC3B versus Myc-BirA*, confirming the isolation of specific LC3B-binding proteins. Several EBV proteins were also identified as predicted LC3B-interactors, BMRF1, BcLF1, BVRF2/BdRF1, BORF2, BXLF1, and BHLF1 (29) (Fig. 2E). To strengthen the validity of these candidates, a second method to identify LC3B-interacting EBV proteins was performed, based on LC3B immunoprecipitation (Fig. 2 F and G). Lysates from EBV(WT) cells, previously induced to the lytic cycle, were immunoprecipitated with either LC3B-specific antiserum or normal rabbit serum (NRS), used as negative control (Fig. 2F), prior to protein identification by MS (Fig. 2G). Similarly, LC3B and LC3B-binding partners, P62, OPTN and ATG3 (30), were found enriched in the LC3B-specific pull-down. Of the four cellular LC3B interactors that we identified during lytic EBV replication (ATG3, P62, optineurin and FYCO1) two (ATG3 and P62) were also identified in the enriched EBV particle fractions (Figs. 1G and 2 E and G). Structural EBV proteins were also enriched: the glycoprotein BZLF2, tegument proteins BGLF2 and BRRF2 and capsid proteins BVRF2/BdRF1 (29). Furthermore, several nonstructural proteins were identified including BaRF1, BORF2, BMRF1, and BBLF2 (Fig. 2G). Interestingly, the EBV proteins BMRF1, BORF2, and BVRF2/BdRF1 were consistently enriched using the two distinct approaches (Fig. 2 E and G) and are strong candidates for interacting with LC3B, possibly participating in the recruitment of autophagy membranes to EBV envelopes. From these, the two capsid proteins BVRF2 and BdRF1 were selected for further studies as potential LC3B interactors, due to their high enrichment in in both MS methods (Fig. 2 E and G) and due to their known structural role during EBV capsid assembly.

Fig. 2.

Bio-ID experiments identify several EBV proteins as potential binding partners of LC3B. (A) Scheme of proximity-dependent biotinylation (Bio-ID) in EBV(WT) cells. EBV(WT) cells stably expressing Myc-BirA*-LC3B will biotinylate proteins that interact or are in proximity with Myc-BirA*-LC3B. After cell lysis, the proteins that carry the biotin tag are isolated by pull-down with streptavidin-coated beads and subjected to MS analysis for identification. (B) Protein biotinylation in WCL of EBV(WT) induced to the lytic cycle (Lytic) and stably expressing Myc-BirA* or Myc-BirA*-LC3B. Cells induced in lytic cycle were incubated with 50 μM biotin for 24 h. After cells lysis, WCL were immunostained for streptavidin, Myc tag and the loading control Vinculin. (C) Infectious EBV particles production expressed in GRU per ml of EBV(WT) and their counterparts stably overexpressing Myc-BirA* or Myc-BirA*-LC3B. Bar graph shows the mean ± SEM of three independent experiments, unpaired t test. (D) Cell lysates from Fig. 2B were submitted to AP using streptavidin-coated agarose beads (AP strep) and the pull-down as well as the supernatant (Sup) were stained with streptavidin to show the enrichment of biotinylated proteins in the pull-down versus Sup. Representative images from two independent experiments. (E) Fold change of proteins identified by Bio-ID and analyzed by MS, comparing Myc-BirA*-LC3B versus Myc-BirA*. White rectangles show LC3B and known LC3B-binding autophagy proteins (positive controls), while black rectangles show the significantly enriched EBV lytic proteins. Bar represents the mean ± STD of two independent experiments. (F) EBV(WT) were induced to the lytic cycle and immunoprecipitated (IP) with LC3B or with NRS (negative control). Pull-downs were immunostained with LC3B to confirm the capture of LC3B before subjected to MS compared to NRS. Representative images from two independent experiments. (G) Fold change of proteins identified by mass spectrometry from LC3B immunoprecipitation versus NRS. White rectangles show LC3B and known LC3B-binding autophagy proteins (positive controls), while black rectangles show the EBV lytic proteins identified. Bar graph represents the mean ± STD of two independent experiments.

The Common Assembly Domain of BVRF2 and BdRF1 Interacts with LC3B.

Several viruses, including poliovirus and coxsackievirus from the Picornaviridae family, as well as herpesviruses VZV, HCMV, and EBV, exploit the autophagy pathway in order to be released in LC3B-positive vesicles (15, 17, 31, 32). However, the viral proteins responsible for the recruitment of LC3B-containing autophagosomal membranes have not yet been identified. In this study, EBV capsid assembly proteins BVRF2/BdRF1 (8) were found associated with EBV enriched EV (Fig. 1E) and were also identified as LC3B interactors (Fig. 2 E and G). Thus, we investigated the interaction between BVRF2/BdRF1 and LC3B during infectious EBV particle production. BVRF2 and BdRF1 are encoded by the same ORF, giving rise to two different transcripts (33) (Fig. 3A). As BdRF1 is contained within the BVRF2 gene (BVRF2/BdRF1), we generated plasmids that encode N-terminal FLAG-tagged versions of either BVRF2/BdRF1 or BdRF1 genes, as well as C-terminal FLAG-tagged BVRF2/BdRF1 (Fig. 3A). EBV(WT) cells were induced into lytic cycle (Lytic) and additionally transfected with either the FLAG-tagged BVRF2/BdRF1 constructs, or a control FLAG-tagged protein (FLAG-CTRL) to investigate their ability to directly bind LC3B (Fig. 3B). The expression of C-terminal FLAG-tagged BVRF2/BdRF1 resulted in two bands corresponding to full length BVRF2 as well as to the autoproteolytically released assembly domain of BVRF2, while several bands were observed for BdRF1-FLAG, the highest one corresponding to BdRF1 (Fig. 3B). Therefore, in these constructs the FLAG tag remains with the shared BVRF2_261-605 domain that is encoded by BdRF1 and autoproteolytically released from the C terminus of BVRF2.

Fig. 3.

The common assembly domain of BVRF2 and BdRF1 interacts with LC3B. (A) Scheme of FLAG-tagged constructs BVRF2/BdRF1 and BdRF1, representing their protein sequence, their specific domains, cleavage sites, and the position of the FLAG tags. (B, Left) WCL of EBV(WT) induced to the lytic cycle (Lytic) and transiently transfected with either FLAG-ALP (FLAG-CTRL), BVRF2/BdRF1-FLAG, or BdRF1-FLAG constructs. Cell lysates were submitted to SDS-PAGE and immunostained for FLAG, LC3B, and the loading control Vinculin. Rectangles represent the protein domain of FLAG-tagged BVRF2/BdRF1 or BdRF1. Right: The same cell lysates were IP using magnetic anti-FLAG-beads. One condition was left without lysate (Beads only) as a negative control. The pulldown from the IP was submitted to SDS-PAGE and immunostained with anti-FLAG and anti-LC3B antibodies. Representative images from three independent experiments. (C, Left) WCL of EBV(WT) induced to the lytic cycle and additionally transiently transfected with GST-LC3B alone, GST-LC3B + FLAG-ALP, GST-LC3B + FLAG-BVRF2/BdRF1 or GST-LC3B + BdRF1-FLAG constructs. The lysates were submitted to SDS-PAGE and stained for FLAG, GST, LC3B, and the loading control Vinculin. Right: The same lysates were IP with GST-coated beads. The pull-down was submitted to SDS-PAGE and immunostained with anti-FLAG and anti-GST antibodies. Representative images from three independent experiments. (D) Scheme of BVRF2 and BdRF1 highlighting the peptides identified in purified EBV (yellow) and LC3B Bio-ID experiments (pink) by MS. Rectangles are scaled to represent the length of the peptides identified. (E) WCL and the corresponding purified EV from EBV(WT) and EBV(WT) cells stably expressing FLAG-BVRF2/BdRF1 were submitted to SDS-PAGE and stained for BNRF1 and FLAG. Representative image from three independent experiments.

The cell lysates were next submitted to FLAG immunoprecipitations, which showed that the FLAG-tagged proteins were successfully isolated (Fig. 3). Only BVRF2/BdRF1-FLAG and BdRF1-FLAG, both giving rise to the FLAG-tagged common assembly domain, were able to specifically pull-down LC3B-I and LC3B-II, while control conditions showed no staining for LC3B. This indicates that the two capsid scaffolding proteins BVRF2 and BdRF1 can bind the two forms of LC3B, with a preference for the membrane-bound form LC3B-II during lytic EBV replication (Fig. 3B) and that their common protein domain (BVRF2_261-605) is responsible for binding LC3B. The interaction of BdRF1 and LC3B was also confirmed in HEK293T cells, which lack the EBV genome and interestingly, BdRF1-FLAG seems to preferentially bind LC3B-I in these cells (SI Appendix, Fig. S3A). Since during the EBV lytic cycle LC3B-II accumulates (17–20), we hypothesize that the preference of BdRF1-FLAG for membrane-bound LC3B during lytic EBV replication might result from this LC3B-II accumulation and from the preferential enrichment of viral capsids at membranes during secondary envelopment.

To verify whether the LC3-binding region is located within BVRF2_261-605, we constructed a plasmid that encodes BVRF2/BdRF1 with a FLAG tag at the N terminus. This construct primarily gives rise to the approximately 30 KDa BVRF2 protease domain that is autoproteolytically released from the N terminus of the full-length BVRF2 with approximately 80 KDa (Fig. 3C). The FLAG-tagged constructs were coexpressed with a GST-tagged LC3B, in order to perform GST immunoprecipitations (Fig. 3C). GST-pulldowns showed that only BdRF1-FLAG is immunoprecipitated with GST-LC3B, and no FLAG bands were observed in the control conditions for FLAG-BVRF2/BdRF1, indicating no interaction of LC3B with the BVRF2 protease domain. These results confirm that the common protein region BVRF2_261-605 is responsible for binding LC3B and thereby interacts with the autophagy machinery (Fig. 3 B and C). Moreover, the binding of BdRF1 with GABARAPL2, another autophagosome protein associated with enriched EBV virions, was also demonstrated, suggesting that BdRF1 also interacts with this LC3 family member involved in autophagosome formation (Fig. 1G and SI Appendix, Fig. S1 F and G)

The location of BVRF2/BdRF1 peptides identified during the MS of the Bio-ID samples (Fig. 2E) corroborate that this shared protein domain binds LC3B (Fig. 3D). Interestingly, BVRF2/BdRF1 peptides detected in the MS of purified EV are spread throughout the entire protein sequence, suggesting that even though the LC3-binding region is located in the assembly domain, the full-length protein could also be associated with extracellular virions (Figs. 1E and 3D). In order to investigate this, FLAG-BVRF2/BdRF1 was stably expressed in EBV(WT) cells and the purified EV enriched in infectious EBV particles produced by these cells were analyzed by western blot. Both FLAG-BVRF2/BdRF1 bands and the viral tegument protein BNRF1 were observed in enriched EBV particles, confirming the presence of the entire BVRF2/BdRF1 protein sequence associated with EBV virions (Fig. 3E). Altogether, these data confirmed that both EBV proteins and ATG proteins can be found in the purified extracellular virions (Figs. 1 and 3 D and E) and identified the BVRF2/BdRF1 common protein region for the direct interaction with LC3B and GABARAPL2 (Fig. 3 A–D and SI Appendix, Fig. S3B). Thus, the capsid scaffolding proteins BVRF2 and BdRF1 seem to be involved in the cross talk with the autophagy pathway via their ability to bind LC3B and GABARAPL2. Indeed, their preferential interaction with the membrane-bound form of LC3B-II together with the association of LC3B-II in virions, suggests that this interaction might contribute to providing autophagic membranes to viral envelopes.

Subcellular Location of BVRF2 and BdRF1 during Lytic EBV Replication.

The capsid assembly proteins BVRF2 and BdRF1 play an important role during capsid assembly in the nucleus (8). BVRF2 is thought to travel to the cytoplasm to interact with the MCP and mediate its translocation to the nucleus to form the capsid scaffold, which has been demonstrated for the homologs of BVRF2 in HSV (UL26) and in HCMV (UL80) (11, 34). However, the subcellular localization of BVRF2 and BdRF1 remains less clear, as their localization has been assessed by overexpressing single proteins without EBV replication, which could influence their subcellular localization and their described roles (9, 29, 35). Additionally, we have shown that the common protein region between BVRF2 and BdRF1 also interacts with LC3B-II (Fig. 3), suggesting that these proteins might play additional roles in the cytoplasm. Thus, we investigated the cellular localization of BVRF2/BdRF1-FLAG and BdRF1-FLAG in EBV(WT) cells induced to the lytic cycle (Fig. 4). First, we noticed by confocal laser scanning microscope that the majority of cells undergoing EBV lytic replication, which were identified by positive VCA staining, showed an accumulation of DNA intercalating dyes such as DAPI at the periphery of the nucleus (Fig. 4A). The VCA punctate staining preferentially localized in the center of the nuclei, associated with low DAPI-fluorescence intensity as shown by the histogram. Indeed, approximately 90% of VCA-positive cells showed alterations in the DAPI signal, and 92% of cells that displayed nuclear reorganization were also positive for VCA staining (Fig. 4A). This characteristic phenotype was also observed by electron microscopy, where 76% of nuclei displaying DNA reorganization were also filled with EBV capsids and similarly, almost 93% of nuclei that had EBV capsids also displayed differences in the nuclear electrodense material (Fig. 4B and SI Appendix, Fig. S4A). Altogether, these nuclear reorganization appear to be a hallmark of cells undergoing lytic EBV replication. Recently, the EBV lytic cycle was described to trigger significant host chromatin modifications (36) and according to our observations, this phenotype could possibly be due to lytic cycle–induced host chromatin rearrangement, which may leave the center of the nucleus devoid of chromatin and instead seems to be filled with EBV capsid proteins. This phenotype offers a marker to identify cells undergoing active EBV replication by fluorescence microscopy. Using the combination of nuclear reorganization phenotype and VCA expression, the subcellular localization of BVRF2 and BdRF1 was assessed in VCA-negative and -positive cells (Fig. 4C).

Fig. 4.

EBV triggers nuclear reorganization, which affects the subcellular localization of BVRF2. (A) EBV(WT) cells induced to the lytic cycle were fixed 72 h postinduction and stained for DAPI (blue) and VCA (red). Histograms show the FI of DAPI and VCA along the white segments in VCA-negative (continuous line) and VCA-positive (dashed line) cells. Bar graphs show the quantitative analysis of percentage of VCA-positive cells that display nuclear reorganization (reorg) and vice versa. Bar graph represents the mean ± SD of three independent experiments, unpaired t test. (B) Electron micrographs of EBV(WT) cells induced to the lytic cycle. Rectangles show enlarged sections, white arrows point to reorganized electrodense material, and white asterisks mark EBV capsids. Quantitative analysis of percentage of capsids with reorganized nuclei that contain EBV capsids and vice versa. Bar graph represents the mean ± SD of two independent experiments. (C–F) EBV(WT) cells lentiviral transduced to express BVRF2/BdRF1-FLAG or BdRF1-FLAG were induced to the lytic cycle and fixed 72 h postinduction. Fixed cells were coimmunostained for DAPI (blue), FLAG (red), and VCA (green). (C) Histograms show fluorescence intensity of DAPI, FLAG, and VCA along the white segments in VCA-positive cells (dashed lines). The approximate location of nuclei and cytoplasmic regions proximal to nuclei are indicated. (D) Quantitative analysis of FLAG signal in nuclei expressing BVRF2/BdRF1-FLAG. Data are normalized to VCA-negative cells. (E) Analysis of Pearson coefficient of colocalization between BdRF1-FLAG and VCA. (F) The ratio of nuclei versus cytoplasmic staining in cells expressing BVRF2/BdRF1-FLAG and BdRF1-FLAG is shown in violin plots. Data representative of four independent experiments, unpaired t test. All confocal images are representative images from three different independent experiments. (Scale bar is 5 µm.)

FLAG staining in BVRF2/BdRF1-FLAG was observed in both cytoplasm and nuclei, as shown by the histogram. However, cells undergoing EBV replication displayed a significant reduction of the nuclear fluorescent intensity (FI) compared to cells that lack VCA staining, which display a strong nuclear signal and a dimmer diffuse cytoplasmic staining (Fig. 4 C and D). This indicates that the nuclear rearrangement triggered by the lytic cycle seems to affect the protein localization of BVRF2/BdRF1-FLAG. Due to its diffuse staining, colocalization of BVRF2/BdRF1-FLAG with punctuate LC3B or Golgi marker Golgin97 appeared unspecific (Fig. 4C and SI Appendix, Fig. S4 B–D). Moreover, we have not been able to observe capsid proteins in the cytoplasm by immunofluorescence nor by EM, as the process of capsid egress from the cells seems challenging to capture by these methods and consequently colocalization of BVRF2/BdRF1 with LC3B was difficult to assess. On the other hand, BdRF1-FLAG displayed almost no staining in the cytoplasm and a strong nuclear FLAG localization in the periphery of the nucleus, which remained unaltered in cells undergoing EBV replication (Fig. 4C). A moderate colocalization of BdRF1-FLAG with the nuclear VCA protein was observed upon EBV replication, as indicated by the Pearson Coefficient of 0.52 (Fig. 4 C and E). These observations are in accordance with the role of BdRF1 as capsid scaffold protein interacting with VCA to form the capsid (8). The quantification of the ratio between nuclei and cytoplasm proximal to nuclei (N/C) confirmed that BVRF2/BdRF1-Flag is also found in the cytoplasm of lytic cells, while BdRF1-FLAG remained mostly nuclear (Fig. 4 C and F). Interestingly, despite the similarities in their protein sequence, the subcellular localization of BVRF2 and BdRF1 was strikingly different, which suggests that BVRF2 could have roles in the cytoplasm, beyond its capsid assembly functions, which could facilitate the interaction with cytoplasmic LC3B-II.

BVRF2 Is an Important Protein for the Generation of Mature Capsids and the Production of Infectious Virions.

BVRF2 and BdRF1 mediate the assembly of EBV capsids in the nucleus but can also interact with LC3B-positive membranes (Fig. 3), strongly suggesting that these proteins may play additional unknown roles during the EBV lytic cycle. As most of the knowledge on BVRF2 and BdRF1 comes from overexpression of these proteins in a nonviral context, we generated a BVRF2 knock-out EBV to further characterize the role of these proteins in ongoing EBV lytic replication (Fig. 5). A mutation in the start codon of BVRF2 was introduced in the EBV WT bacmid, abolishing the expression of BVRF2 (Fig. 5A). Nevertheless, the protein BdRF1 could still be expressed although it remains uncleaved, due to the absence of BVRF2 and its protease activity. The successful mutation was confirmed by sequencing (Fig. 5B), and the integrity of the bacmid was verified by digesting the Bacmid with BamHI and BgLI and comparing the fragments to the WT bacmid (Fig. 5C). The ∆BVRF2 bacmid was transfected into HEK293 cells to generate a stable cell line, [EBV(ΔBVRF2)] used to investigate the role of BVRF2 in the EBV life cycle. It is important to note that the EBV(ΔBVRF2) cell line has been subcloned to a different degree than EBV (WT) cells, which have been optimized to maximize virus production. Thus, EBV(WT) cells produce 10-fold higher EBV titers than BVRF2 transcomplemented EBV(ΔBVRF2) cells and are used as a positive technical control but cannot directly be compared to EBV(ΔBVRF2). The ability of EBV(ΔBVRF2) cells to produce infectious virus was first assessed in the absence of BVRF2 or upon the transcomplementation of FLAG-BVRF2/BdRF1 or BdRF1-FLAG (Fig. 5D). EBV(ΔBVRF2) cells showed a ten-fold increase in EBV production upon induction to the lytic cycle compared to control condition. However, this increase merely reached 100 GRU/mL, while the transcomplementation of BVRF2/BdRF1 rescued the level of produced EBV infectious particles to 10,000 GRU/mL. Interestingly, transcomplementation of BdRF1-FLAG did not rescue the virus production, as the levels of infectious EBV were in the same range as the EBV(∆BVRF2) lytic condition. This result indicates that BVRF2 is an important protein for the production of infectious virions (Fig. 5D). As EBV(∆BVRF2) cells had an impairment in producing infectious EBV, we further investigated which step of virus assembly is affected by the absence of BVRF2. First, we assessed if the nuclear reorganization characteristic for the EBV lytic cycle (Fig. 4 A and B) was altered by the lack of BVRF2 expression (Fig. 5E). We found that 92% of EBV(∆BVRF2) cells in lytic stage, indicated by the expression of VCA, displayed a nuclear reorganization, and vice versa; 89% of nuclei with altered DAPI signal were positive for VCA staining, which was comparable to the frequencies observed in EBV(WT) cells (Figs. 5E and 4 A and B). Therefore, BVRF2 seems not to be involved in the development of nuclear reorganization in virus-producing cells. We next investigated the morphology of the EBV capsids, produced by EBV(∆BVRF2) cells, using transmission electron microscopy (TEM) (Fig. 5F). EBV(WT) nuclei analyzed by TEM displayed a mixture of EBV capsids in different stages of maturation, including capsids without an electrodense core (procapsid) (8) and mature capsids, which represent 47.5% of the analyzed capsids. In contrast, cells lacking BVRF2 were unable to form mature capsids, as only procapsids were observed in nuclei of EBV(∆BVRF2) cells (Fig. 5F). Taken together, these data demonstrated that BVRF2 is an important protein for the assembly of mature EBV capsids and the production of infectious virions.

Fig. 5.

BVRF2 is required for the generation of mature capsids and plays a major role in the generation of infectious virions. (A) Scheme of the cloning performed on WT p2089 bacmid, in which the ATG start codon of BVRF2 was exchanged for a RpsL/Kanamycin cassette. This cassette was then substituted for the TAA codon, whereby the transcription of BVRF2 is abrogated. (B) DNA Sanger sequencing on the WT p2089 and ∆BVRF2 p2089 bacmid where asterisks indicate the nucleotides mutated. (C) Picture from the agarose gel showing the digestion of the three bacmids by BamHI and BgLI restriction enzymes and confirming the integrity of the final ∆BVRF2 p2089 bacmid. (D) Infectious EBV particles production expressed in GRU per mL of EBV(WT) or EBV(∆BVRF2) cells induced to the lytic cycle. EBV(∆BVRF2) cells were also transcomplemented with FLAG-BVRF2/BdRF1 or BdRF1-FLAG. Scatter plot shows mean ± SEM of six independent experiments, unpaired t test. (E) EBV(∆BVRF2) cells induced to the lytic cycle were fixed 72 h postinduction and stained for DAPI (blue) and VCA (red). Confocal images are representative images from three different independent experiments. (Scale bar is 5 µm.) Histograms show the FI of DAPI and VCA along the white segments in VCA-negative (continuous line) and -positive cells (dashed line). Quantitative analysis of percentage of VCA-positive cells that display nuclear reorganization (reorg) and vice versa. Bar graph represents the mean ± SEM of three independent experiments, unpaired t test. (F) Electron micrographs of EBV(WT) and EBV(∆BVRF2) cells induced to the lytic cycle. Rectangles show enlarged sections, displaying high-resolution capsid images. Quantitative analysis of the percentage of mature capsids, with at least 7 capsids quantified per cell. Bar graph represents mean ± SD of three independent experiments, unpaired t test.

Both BVRF2 and BdRF1 as Well as the Lipidation of LC3B Play a Role in the Secretion of Viral Vesicles.

To further investigate the relevance of BVRF2 as an important protein during EBV lytic cycle, we tested the ability of EBV(∆BVRF2) cells to sustain viral DNA replication (Fig. 6A). As expected, EBV(WT) cells showed a significant increase in viral DNA upon the induction of the lytic cycle and were used as a positive control. Similarly, EBV(∆BVRF2) cells were also able to replicate viral DNA, shown by the significant 12-fold increase in EBV DNA copies. The transcomplementation of BdRF1-FLAG showed a 22-fold increase in viral DNA copies, while transcomplementation of FLAG-BVRF2/BdRF1 significantly increased the replication of EBV DNA by 28-fold compared to control conditions (Fig. 6A). Therefore, these data indicate that BVRF2 is not required for the replication of EBV genomes, although its overexpression seems to confer an advantage to the virus by boosting the DNA copy numbers. Next, to assess whether BVRF2 could play a role in the transcriptional cascade of EBV lytic cycle, we analyzed the expression of EBV early lytic protein BMRF1 and late capsid protein P18 in EBV(∆BVRF2) cells (Fig. 6B). Both EBV(WT) cells as well as EBV(∆BVRF2) expressed BMRF1 and P18 when induced to lytic cycle and also during trans-complemented conditions with FLAG-BVRF2/BdRF1 or BdRF1-FLAG expression. This demonstrates the successful induction of the lytic cycle but most importantly, that the lack of BVRF2 does not influence the expression of EBV proteins (Fig. 6B). The quantification of the band intensity of BMRF1 and P18 showed no significant differences (Fig. 6B) when BVRF2/BdRF1 or BdRF1 are expressed in EBV(ΔBVRF2) cells, indicating that BVRF2 does not play a role in the efficient transcription and translation of EBV lytic proteins. Whether BVRF2/BdRF1 or BdRF1 influences the autophagy pathway was analyzed by overexpressing these proteins in EBV(WT), EBV(∆BVRF2) and HEK293T cells (SI Appendix, Fig. S5 A–C). While the expected increase in LC3B-II was observed upon the induction of the lytic EBV cycle, overexpression of BVRF2/BdRF1-FLAG nor of BdRF1-FLAG did not lead to an increase in LC3B-II in the presence or absence of lytic EBV replication as well as in HEK293T cells that do not carry the EBV bacmid. Altogether, we have observed that in the absence of BVRF2, EBV capsids were unable to mature and form infectious virions (Fig. 5). In addition, as viral components did not seem to accumulate in cells (Fig. 6B), we hypothesized that noninfectious viral particles might still assemble and be released. To measure the egress of viral vesicles (infectious viruses, virus-like particles, and viral envelopes), we performed attachment assays, which allow the detection and quantification of GP350-containing vesicles by flow cytometry (Fig. 6C). First, the ability of both EBV(WT) and EBV(∆BVRF2) to express GP350 was assessed by immunofluorescence, and we observed that, in both cell lines, more than 90% of VCA-positive cells also showed staining for GP350 (SI Appendix, Fig. S6A). This demonstrates that the expression of BVRF2 does not influence the expression of GP350. Next, the number of GP350-positive vesicles was measured in lytic-induced EBV(∆BVRF2) cells transcomplemented with FLAG-BVRF2/BdRF1 or FLAG-BdRF1 (Fig. 6D). Supernatants from EBV(WT) cells were used as positive control and noninduced EBV(∆BVRF2) cells were used as negative control. The lytic induction of EBV(∆BVRF2) showed a threefold increase in GP350 mean FI (MFI), indicating that GP350 vesicles can be released in absence of BVRF2. The transcomplementation of BVRF2 did not further increase the release of GP350 vesicles, as it showed a significant 3.3-fold increase of GP350 MFI, even so this condition represents the maximal infectious EBV particle production that this cell line can perform. Transfection of BdRF1 resulted in a 2.1-fold increase compared to UI cells (Fig. 6D). These data suggest that BVRF2 deficiency does not decrease the release of viral vesicles, but, as both BVRF2 and BdRF1 are able to bind LC3B (Fig. 3), BdRF1 might still bind LC3B and mediate the viral envelope egress in the EBV(∆BVRF2) cells.

Fig. 6.

Both BVRF2/BdRF1 and LC3 lipidation play a major role in the secretion of viral vesicles. (A–D) EBV(WT) and EBV(∆BVRF2) cells were induced to the lytic cycle (Lytic) and transcomplemented with FLAG-BVRF2/BdRF1 or BdRF1-FLAG for 72 h. (A) Extracted DNA was submitted to qPCR targeting BMRF1 gene, and data were normalized to the respective UI condition. Data are expressed in fold change in EBV DNA copy number. Scatter plot shows mean ± SEM from four independent experiments. Statistical analysis: unpaired t test. (B) Cell lysates were submitted to SDS-PAGE and stained for FLAG, early lytic EBV protein BMRF1, late lytic capsid protein P18 and loading control protein Vinculin. P18 and BMRF1 bands in EBV(∆BVRF2) were quantified relative to Vinculin and normalized to the lytic condition. Bar graph shows mean ± SD from six independent experiments, unpaired t test. (C) Scheme of ELIJAH assay. ELIJAH B cells are incubated with supernatants from cells induced to the lytic cycle, so that viral GP350-containing vesicles attach to the complement receptor 2 (CD21) on B cells. Next, staining with fluorescently labeled anti-GP350 antibodies allows the detection by flow cytometry of cells carrying infectious EBV as well as noninfectious GP350-positive vesicles. (D) Measurement of GP350⁺ vesicles secreted by EBV(WT) and EBV(∆BVRF2) cells via ELIJAH assay. Data expressed in MFI normalized to the CTRL condition. Scatter plot represents mean ± SD from four independent experiments, unpaired t test. (E) EBV(∆BVRF2) cells were silenced with shRNA against SCMBL, BdRF1, ATG5, ATG12, and ATG16L1. Supernatants were analyzed by ELIJAH assay to measure secretion of GP350⁺ vesicles. Data expressed in MFI normalized to shSCMBL condition, and dotted line represents the average of CTRL condition. Cell lysates were submitted to SDS-PAGE and stained for EBV proteins BMRF1 and P18 and the loading control Vinculin. Scatter plot represents mean ± SEM from five independent experiments, one-way ANOVA.

Thus, to test whether both LC3B-interacting proteins BVRF2 and BdRF1 are required for viral envelope release, we silenced the common assembly domain of both proteins (shBdRF1) in EBV(∆BVRF2) cells to impair the expression of both proteins, or silenced the members of the E3-like LC3B lipidation complex ATG5-ATG12-ATG16L1, required to produce LC3B-II (Fig. 6E). The efficient induction of the lytic cycle was confirmed by the expression of lytic proteins BMRF1 and P18 (Fig. 6E). The successful silencing of ATG5, ATG12, and ATG16L1 proteins was demonstrated by western blotting (SI Appendix, Fig. S6B). The lentivirus construct for silencing BdRF1 was validated after overexpression of FLAG-tagged BVRF2/BdRF1 and BdRF1 (SI Appendix, Fig. S6C). Control silenced EBV(∆BVRF2) cells (shSCMBL) that were additionally induced to enter the lytic cycle showed a significant increase in the production of viral envelope-like GP350⁺ vesicles compared to CTRL conditions (Fig. 6E). Upon silencing of BdRF1 in the absence of BVRF2, the secretion of viral vesicles was significantly decreased to levels of the UI condition. Silencing of ATG5, ATG12, and ATG16L1 also led to a comparable reduction in GP350⁺ vesicles, indicating that both BVRF2/BdRF1 and the autophagy machinery, responsible of generating LC3B-II, are required for the secretion of viral vesicles. These data indicate that the LC3B-interacting viral proteins BVRF2 and BdRF1 as well as the LC3B conjugation complex of ATG5-ATG12-ATG15L1 contribute to the release of GP350-containing viral vesicles, likely due to BVRF2/BdRF1 recruiting LC3B-containing membranes that then can be found in infectious mature virions.

Discussion

Several other herpesviruses also manipulate autophagy during their lytic cycle and use autophagic membranes as a source for their envelope (21). Consequently, proteins of the autophagy machinery and of endosomal compartments have been found associated with extracellular virions. Indeed, the β-herpesvirus HCMV carries the autophagy proteins ATG5, BECN1, P62, GABARAP as well as LC3B-II in its infectious virus particles (15, 37). The α-herpesvirus VZV contains LC3B and the marker of recycling endosomes RAB11 in its virions (32). We could now identify components from both the BECN1 containing PI3 kinase complex, the LC3 lipidation complex, LC3B-II, and GABARAPL2, as well as autophagy receptors in EBV particles. EBV seems to stabilize autophagic membranes during lytic replication (17, 20), possibly via the early lytic protein BFRF1 (38). In this study, we show that these stabilized LC3B-coated autophagic membranes might be recruited by BVRF2/BdRF1 binding to LC3B-II into the EBV envelope and thereby end up in the released infectious virus particles. Without BVRF2/BdRF1 or in the absence of the LC3B conjugation complex of ATG5-ATG12-ATG15L1, viral envelope release is compromised.

In addition to herpesviruses, several RNA viruses also seem to use autophagic membranes for their infectious virus production. Autophagic membrane reorganization has been observed during influenza, picornavirus, flavivirus, and coronavirus replication (39–47). Some of these RNA viruses use autophagic membranes to establish their perinuclear replication compartments, from which picornaviruses like polio and coxsackie B virus travel along microtubules in a LC3-dependent fashion for nonlytic release (48, 49). Autophagic membranes seem to be stabilized for this picornavirus secretion by inhibition of the fusion machinery of autophagosomes with lysosomes (50). Furthermore, viral protein targeting for degradation by autophagy receptors like P62 seems to be inhibited by viral protease-mediated cleavage of these receptors (51, 52). This stabilization of autophagic membranes and inhibition of autophagic degradation of viral proteins results in the release of packages of the nonenveloped picornaviruses in LC3-II-decorated membranes (31, 53, 54). This mechanism requires the PI3 kinase and the LC3 lipidation complexes, but not the ULK1-associated protein kinase complex (55). The LC3-II-enveloped packages of 350 to 500 nm in diameter contain 20 to 30 picornavirus particles and represent an alternative transmission strategy for these nonenveloped RNA viruses. Exactly the autophagy components that are required for nonlytic picornavirus release are also present in enriched EBV particles in this study, suggesting similar mechanisms to subvert the autophagy pathway for enhanced viral egress.

Sampling autophagic membranes into viral envelopes or membranes surrounding multiple viral particles might have several advantages for the respective viruses, including EBV (56, 57). The autophagy machinery can recruit membranes from several organelle sources but mainly assembles autophagosomes at the endoplasmic reticulum (58) or uses membranes of multivesicular bodies for extracellular vesicle release (59). These membrane sources enriched in phosphatidylserine are advantageous for the virus, as the incorporation of these lipids into the outer membrane leaflet of viral envelopes can be used by scavenger receptors for uptake by new host cells (53). For EBV, membrane-bound viral glycoproteins, such as GP350, are recruited to the viral envelope before its BVRF2/BdRF1 and LC3 lipidation dependent release. Such a release is nonlytic and therefore prevents the death of the virus-producing cells, even so this harmless egress is more important for nonenveloped viruses, like picornaviruses (49). Once released, envelopes with autophagic membranes protect from capsid and tegument specific antibodies, which are also abundantly produced during EBV infection (60). Moreover, enveloped viruses or virus particle packages include cellular factors such as proteins and nontranslated RNAs to modify host cell biology upon infection (61). Membrane recruitment via the autophagy machinery for exocytosis of EV has even been described to preferentially incorporate RNA and RNA binding proteins (62). The alternative use of the autophagy machinery for exocytosis after inhibition of lysosomal degradation, observed during lytic EBV replication (17, 20), could be linked to the release of autophagy receptors in EV (63), such as SQSTM1/P62. Accordingly, we detected autophagy receptors SQSTM1/P62 as well as NBR1 in enriched EBV particles. Thus, some herpes and picornaviruses might hijack extracellular vesicle release pathways that utilize the autophagy machinery to recruit their envelopes or membranes around viral packages. This appears beneficial to escape antibody-mediated immunity and to deliver cellular factors, viral proteins, and genomes during infection of new host cells.

Materials and Methods

A complete description of the materials sources, reagents, and our methods can be found in SI Appendix, SI Materials and Methods. It includes a description of the generation of EBV(ΔBVRF2) cells; virus production, concentration and purification; mass spectrometry, immunofluorescence and immunoprecipitation procedures; as well as a table of reagents and statistical analyses.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (PDF)

Dataset S02 (PDF)

Dataset S03 (PDF)

Dataset S04 (PDF)

Dataset S05 (XLSX)

Dataset S06 (XLSX)

Dataset S07 (XLSX)

Data, Materials, and Software Availability

The mass spectrometry reported in this paper has been deposited in the b2share database, https://doi.org/10.23728/b2share.1f9707032b4a43a3ae7d903cf1edf94e (64). All other data are included in the article and/or supporting information.