Вирус Эпштейн-Барр BALF0/1 подрывает пути кавеолина и ERAD, чтобы охватить комплексы рецепторов клеток B для разложения.

Epstein–Barr virus (EBV) establishes lifelong infection in >95% of adults worldwide, is the major viral trigger for multiple sclerosis, and is associated with ~2% of human cancers ( 1 , 2 ). These include endemic Burkitt lymphoma, Hodgkin lymphoma, posttransplant and HIV-associated lymphomas, T and NK-cell lymphomas, and nasopharyngeal and gastric carcinomas ( 3 , 4 ). EBV uses a biphasic lifecycle to colonize the B cell compartment, to spread between B and epithelial cells, and to transmit between hosts ( 5 – 7 ). The EBV lytic cycle is triggered by the immediate early genes BZLF1 and BRLF1, which induce expression of 35 early genes important for viral DNA replication and then 40 late genes that include structural proteins important for production of infectious virions. Much remains to be learned about how EBV subverts host innate and adaptive immune barriers in order to establish latency, reactivate within the heart of the adaptive immune system.

EBV has evolved multiple mechanisms to interfere with antigen presentation pathways, thereby suppressing cell-mediated immune responses ( 8 – 15 ). However, elevated antibody titers against viral antigens are found in numerous EBV-associated disease states, including with infectious mononucleosis, endemic Burkitt lymphoma, nasopharyngeal carcinoma, classical Hodgkin lymphoma, and multiple sclerosis ( 16 – 18 ). While Epstein–Barr virion evasion of complement activation has been described ( 19 ), viral strategies to protect virions from immunoglobulin, including that produced within the infected cell where viral particles mature, are less well-studied.

To gain insights into how EBV remodels host B cells in support of lytic replication, we used tandem-mass-tag-based MS3 mass spectrometry to perform quantitative temporal proteomic analysis in EBV+ Burkitt lymphoma B cells, prior to and at four time points following lytic reactivation ( 20 ). Unexpectedly, this analysis revealed that B cell receptor (BCR) complexes, composed of immunoglobulin (Ig) heavy chain (HC), light chain (LC), and noncovalently associated CD79A/B (also known as Igα/β) signaling chains, were selectively and highly depleted upon EBV lytic reactivation. In IgM-expressing cells, the secretory IgM joining-chain (J-chain) was also depleted, whereas only a small number of host cell proteins exhibited significant downmodulation, suggesting selective posttranslational targeting ( 20 ). Whereas activated BCR complexes are typically delivered to lysosomes to support presentation of peptide antigens to CD4+ T cells by major histocompatibility complex (MHC) class II molecules ( 21 ), EBV instead routes BCR to cytosolic proteasomes by an unidentified pathway ( 20 ). Which EBV gene targets BCR complexes for degradation, and how it targets BCR complexes embedded within B cell membranes to cytosolic proteasomes has remained unknown. However, acyclovir treatment did not prevent EBV lytic cycle BCR depletion, indicating that viral late lytic genes are not required ( 20 ). Here, we used a recently constructed EBV lytic cycle protein–protein interaction map ( 22 ) to generate the hypothesis that the EBV early lytic cycle encoded BALF0/1 targets BCR complexes for proteasomal degradation. We present evidence that BALF0/1 subverts caveolar endocytosis and ER-associated degradation pathways to target BCR for degradation.

Results

BALF0/1 Targets the B Cell Receptor for Degradation During EBV Lytic Cycle.

To gain insights into how EBV targets the BCR for cytoplasmic proteasome degradation, we interrogated a recently generated EBV protein–protein interaction network map, generated in Burkitt lymphoma cells induced for EBV lytic replication ( 22 ). Our proteomic analysis identified high-confidence interactions between the immunoglobulin (Ig) HC and two EBV early lytic proteins, BALF0/1 and BARF1 ( Fig. 1 A ). BALF0/1 associated with multiple proteasome components, with proteins involved in microtubule associated protein transport, and with B cell receptor-associated protein 31 (BCAP31), a chaperone that plays a role in the ER associated degradation (ERAD) pathway ( Fig. 1 A ). Of note, two in-frame methionine codons can be used to generate 220 residue BALF0 or 189 residue BALF1 ( 23 ). AlphaFold modeling predicts that BALF0/1 and the antiapoptotic BCL2 family member MCL1 share structural homology ( SI Appendix , Fig. S1 A ). Similarly, proteomic analysis highlighted interaction between BARF1 and several ubiquitin–proteasome pathway proteins ( Fig. 1 A ) ( 22 ). BARF1 exhibits homology with the colony-stimulating factor receptor and also has limited amino acid sequence similarity with CD80 ( 11 , 24 ).

EBV early lytic protein BALF0/1 targets the B cell receptor complex for degradation. ( A ) EBV B cell lytic cycle BALF0/1 ( Left ) and BARF1 ( Right ) protein–protein interaction network, adapted from ref. 22 . DAVID was used to identify enriched pathways among BALF0/1 interactors. ( B ) BALF0/1 vs. BARF1 effects on BCR abundance. Immunoblot analysis of IgM HC and CD79B abundances in whole cell lysates (WCL) from latent P3HR-1 Burkitt cells induced for expression of BALF0/1 or BARF1 cDNAs by 5 µg/mL doxycycline (Dox) for 24 h, as indicated. * denotes bands for BALF0, BALF1, or BARF1. 12% SDS-PAGE gel was used for separation of BALF0 and BALF1 proteins. ( C ) BALF0 vs. BALF1 effects on BCR abundance. Top : Immunoblot analysis of IgM HC and CD79A abundances in WCL from P3HR-1 Burkitt cells induced for expression of BALF0 or BALF1 cDNAs by 5 µg/mL Dox for 24 h, as indicated. * denotes bands for BALF0 or BALF1. 12% SDS-PAGE gel was used for separation of BALF0 and BALF1 proteins. Bottom : Schematic diagram illustrating the BALF0 vs. BALF1 open reading frames. ( D ) Analysis of whether BXLF1 vs. BALF0/1 are required for EBV lytic cycle BCR targeting. Immunoblot analysis of WCL from Cas9+ P3HR-1 Z-HT/R-HT cells expressing the indicated sgRNA, induced into lytic replication by 400 nM 4-HT for 24 h. ( E ) Mean ± SEM percentage of cells with detectable IgM signal as in SI Appendix , Fig. S1 F of P3HR-1 cells expressing the indicated BXLF1 or BALF0/1 sgRNA, 4-HT induced as indicated, using data from 15 randomly selected panels of 300 cells from n = 3 replicates, analyzed using the ImageJ ComDet plugin. Statistical analysis was performed with Student’s t test unless otherwise specified. *** P < 0.001. See also SI Appendix , Fig. S1 . Blots in B – D are representative of at least n = 2 replicates.

To test whether either BALF0/1 or BARF1 can independently deplete BCR components, we conditionally expressed cDNA encoding either of these two EBV early genes in P3HR-1 Burkitt lymphoma cells. Expression of BALF0/1, but not BARF1, depleted both CD79B and IgM HC, suggesting that BALF0/1 can trigger BCR depletion even in the absence of an EBV lytic cycle ( Fig. 1 B and SI Appendix , Fig. S1 B ). We next asked whether both BALF0 and BALF1 can each target BCR for degradation. To express only BALF0, we derived a BALF0 methionine 39 to alanine point mutant to prevent expression of BALF1, whose initiator codon is methionine 39. To express BALF1, we instead expressed an N-terminal deletion mutant lacking the first 38 BALF0 residues. Expression of either was sufficient to deplete IgM HC and CD79A ( Fig. 1 C and SI Appendix , Fig. S1 C ).

We next measured BCR subunit abundances in BALF0/1+ P3HR-1 cells, in the absence or presence of the proteasome inhibitor bortezomib ( 20 ). Bortezomib increased IgM HC abundance in BALF0/1 expressing cells ( SI Appendix , Fig. S1 D ). To build on these results, we next asked whether BALF0/1 was necessary for BCR loss in B cells triggered for lytic reactivation. For this analysis, we used Cas9+ P3HR-1 cells with a conditional EBV lytic reactivation system, where the EBV immediate early BZLF1 and BRLF1 alleles are fused to a mutant estrogen receptor binding domain ( 25 , 26 ). Addition of 4-hydroxytamoxifen (4-HT) triggers nuclear translocation of the BZLF1 and BRLF1 chimeras, termed Z-HT and R-HT, respectively, to drive EBV early gene expression and the viral lytic cycle. To establish CRISPR-edited cells, single guide RNAs (sgRNAs) were separately expressed to target BALF0/1 or instead as a control to target the EBV early lytic gene BXLF1, which encodes the viral thymidine kinase that is dispensable for EBV replication ( 27 , 28 ). BALF0/1 or BXLF1 knockout (KO) did not perturb expression of the endogenous immediate early BZLF1 or early BMRF1 genes upon 4-HT addition, indicating fidelity of EBV genomes in CRISPR-edited P3HR-1 cells. Consistent with overexpression analysis, CRISPR editing of BALF0/1, but not of control BXLF1, significantly increased IgM HC and CD79B abundances in lytic P3HR-1 cells ( Fig. 1 D and E and SI Appendix , Fig. S1 E and F ). Furthermore, we validated that IgM HC coimmunoprecipitated with BALF0/1, but not with control GFP from whole cell lysates of bortezomib-treated P3HR-1 cells induced for lytic replication ( SI Appendix , Fig. S1 G ). Collectively, our results suggest that BALF0/1 associates with and destabilizes BCR complexes, beginning in the early phase of EBV lytic replication.

BALF0/1 Utilizes the Caveolae Pathway to Internalize Plasma Membrane BCR.

We investigated how cytoplasmic BALF0/1 routes membrane-embedded BCR complexes, including the large pool of plasma membrane (PM) BCR, to cytoplasmic proteasomes. We hypothesized that BALF0/1 must first trigger BCR endocytosis prior to its degradation, given that we previously observed intracellular IgM HC in bortezomib-treated cells triggered for lytic replication ( 20 ). To further characterize this process, we performed confocal microscopy on 4-HT and bortezomib-treated P3HR-1 cells expressing control BXLF1 or BALF0/1 sgRNAs. Consistent with our prior observations, IgM puncta were evident in bortezomib-treated lytic BXLF1 KO control cells, but had a distinct distribution in BALF0/1-edited cells that resembled the pattern in uninduced cells ( SI Appendix , Fig. S2 A ). To further characterize IgM distribution in BALF0/1-depleted cells, we costained IgM and the PM in BXLF1 control vs. BALF0/1 sgRNA expressing cells reactivated by 4-HT and also treated with bortezomib to block proteasomal degradation. IgM signal was largely distinct from PM stain in control cells, but overlapped in BALF0/1-depleted cells ( Fig. 2 ). Moreover, BALF0/1 colocalized with internalized IgM in many bortezomib-treated P3HR-1 cells induced for BALF0/1 expression and with IgM in rare cells, in which IgM was not depleted in the absence of bortezomib treatment ( SI Appendix , Fig. S2 B ). As BALF0/1 is not known to insert into the PM, this result suggests that it triggers BCR endocytosis.

BALF0/1 triggers endocytosis of the BCR complex. Analysis of BCR subcellular distribution in BXLF1 vs. BALF0/1 edited, bortezomib-treated lytic cells. Immunofluorescence analysis of PM and IgM subcellular distribution in Cas9+ P3HR-1 cells expressing the indicated BXLF1 targeting or BALF0/1 targeting sgRNA. Cells were 4-HT induced for lytic replication and treated with bortezomib for 24 h. Shown beneath are fluorescence intensity line scanning of the regions indicated by the white diagonal dotted line within the boxed area for PM (red) and IgM (green). See also SI Appendix , Fig. S2 .

Caveolar and clathrin-mediated endocytosis are the two major pathways that internalize PM receptors. Clathrin pathway cargo are predominantly recycled or routed to lysosomes for degradation ( 29 ), whereas caveolar pathway cargo travel retrograde to the Golgi, ER, or to endosomes ( 30 – 34 ). Since our proteomic analysis also identified high confidence BALF0/1 associations with proteins related to calmodulin and microtubule motor activity ( Fig. 1 A ), we next investigated whether either pathway was required for BALF0/1 downmodulation of BCR complexes. Inhibition of the clathrin endocytosis pathway by chlorpromazine ( 35 , 36 ) did not rescue BCR expression in lytic B cells ( SI Appendix , Fig. S3 A and B ). By contrast, inhibition of the caveolae-dependent endocytosis by genistein ( 37 ) restored BCR expression to a similar degree as in bortezomib treatment in P3HR-1 in the presence of BALF0/1 ( SI Appendix , Fig. S3 C and D ). Moreover, CRISPR depletion of the key caveolae structural protein caveolin-1 (CAV-1), which is required for caveolar internalization, also impaired IgM depletion upon EBV lytic reactivation ( Fig. 3 A and SI Appendix , Fig. S3 E ). Furthermore, IgM remained largely colocalized with the PM in CAV-1-depleted cells ( Fig. 3 B ). These results suggest that BALF0/1 triggers PM BCR internalization via the caveolin pathway.

BALF0/1 triggers BCR endocytosis via the caveolae pathway. ( A ) Dependency of BALF0/1-mediated BCR turnover on CAV-1. Immunoblot analysis of WCL from P3HR-1 cells expressing control or caveolin-1 (CAV-1) targeting sgRNA and 4-HT induced into lytic replication for 24 h, as indicated. Blots are representative of ≥n = 2 replicates. See also SI Appendix , Fig. S3 E . ( B ) Immunofluorescence analysis of PM, IgM, and calnexin in Cas9+ P3HR-1 cells expressing CAV-1 targeting sgRNA, 4-HT induced into lytic cycle for 24 h. ( C ) Immunofluorescence analysis of IgM vs. calnexin subcellular distribution in P3HR-1 cells expressing control or CAV-1 targeting sgRNA, induced for lytic replication by 4-HT treatment for 24 h. ( D ) Mean ± SEM numbers of colocalized CAV-1 and IgM puncta per P3HR-1 cell (as in panel C ) expressing the indicated control or CAV-1 sgRNA, 4-HT induced, as indicated, using data from 12 randomly selected panels of 240 cells from n = 3 replicates, analyzed using the ImageJ ComDet plugin. Statistical analysis was performed with Student’s t test unless otherwise specified. *** P < 0.001. See also SI Appendix , Fig. S3 .

Consistent with a major CAV-1 role downstream of BALF0/1, EBV lytic reactivation triggered extensive CAV-1 puncta formation in control BXLF1 KO, but to a lesser extent in BALF0/1 KO P3HR-1 triggered for reactivation by 4-HT ( Fig. 3 C and D ). EBV lytic reactivation triggered by IgG crosslinking similarly induced CAV-1 puncta, which colocalized with IgG in control BXLF1 edited cells to a greater degree than in BALF0/1 edited cells ( SI Appendix , Fig. S3 F and G ). IgG cross-linking triggered CAV-1 puncta that more frequently colocalized with IgG in EBV+ Akata cells than in an EBV- Akata subclone ( 38 ) ( SI Appendix, Fig. S3 H ). These results suggest that the EBV lytic cycle, rather than IgG cross-linking alone, induces IgG internalization into CAV-1-containing vesicles at the 48 h timepoint post-Ig crosslinking.

BALF0/1 Uses the ER-Associated Degradation Pathway to Route BCR to Proteasomes.

How might BALF0/1 route membrane-bound BCR complexes to cytosolic proteasomes? Given the observed dependency on caveolae, we hypothesized that BALF0/1 may utilize a similar pathway to that subverted by cholera toxin, which following caveolar uptake, travel retrograde to the ER, from where they are routed to the cytosol ( 39 ). Similarly, misfolded cargo are routed from the ER to cytosolic proteasomes via the ER-associated degradation (ERAD) pathway ( 40 – 42 ). To test this, we performed confocal microscopy on latent vs. reactivated P3HR-1, in the absence or presence of bortezomib to stabilize BCR in reactivated cells. 4-HT triggered a significant degree of colocalization between relocalized IgM HC and the ER-resident chaperone calnexin in bortezomib-treated cells, which were largely distinct from the PM ( SI Appendix , Figs. S4 A and B and S5 A – C ). Similar calnexin/IgG colocalization was observed in bortezomib-treated EBV+ Akata cells triggered for lytic reactivation by IgG-crosslinking ( SI Appendix , Fig. S5 D – F ). We note that not all 4-HT or anti-IgG treated cells reactivate EBV, and that not all BCR foci of aggregation were captured in a two-dimensional slice, potentially resulting in an underestimate of the extent of EBV lytic cycle induced Ig/calnexin colocalization. These results suggest that BALF0/1 routes internalized BCR to the ER, and that proteasome inhibition causes a BCR population to be retained within the ER.

To examine whether BALF0/1 routes internalized BCR from the ER to cytosolic proteasomes, we depleted HRD1 (HMG-CoA reductase degradation protein 1), an ER-resident E3 ubiquitin ligase that forms an ER-membrane spanning channel important for ERAD ( 43 – 45 ). HRD1 KO markedly diminished the extent of IgM and CD79A loss observed in P3HR-1 induced for lytic reactivation, despite having no appreciable effects on immediate early BZLF1 or early gene BMRF1 expression ( Fig. 4 A and SI Appendix , Fig. S6 A ). Similarly, the small-molecule ERAD inhibitor eeyarestatin I ( 46 ), which blocks AAA ATPase p97/VCP mediated cytoplasmic shuttling of ERAD cargo ( 47 ), significantly impaired BCR degradation in a dose-dependent manner, albeit with a degree of EBV lytic cycle gene repression. By contrast, inhibition of lysosomal protease activity by leupeptin, E64, or bafilomycin failed to rescue EBV lytic cycle BCR degradation ( SI Appendix , Fig. S6 B and C ). Also suggestive of an ERAD role in BCR turnover, we observed BCR aggregates that colocalized with calnexin in a significant number of P3HR-1 HRD1 KO cells induced for lytic replication, which accumulated to a similar degree as in bortezomib-treated cells ( Fig. 4 B and SI Appendix , Fig. S6 D – F ). In reactivated P3HR-1, eeyarestatin caused intracellular IgM accumulation. This intracellular IgM colocalized with calnexin, but not PM signal ( SI Appendix , Fig. S7 A ). Despite their involvements in some but not all ERAD pathways, depletion of either BCAP31 or the E3 ubiquitin ligase PARK2 did not noticeably stabilize IgM or CD79A in Burkitt cells triggered for EBV reactivation, suggesting that they are not required for BCR degradation downstream of BALF0/1 ( SI Appendix , Fig. S7 B – E ).

BALF0/1 subverts the ER-associated degradation pathway to route BCR to cytosolic proteasomes. ( A ) Analysis of BALF0/1-mediated BCR degradation dependency on the ERAD pathway ubiquitin ligase HRD1. Immunoblot analysis of WCL from Cas9+ P3HR-1 cells expressing the indicated HRD1-targeting sgRNA, 4-HT induced into lytic cycle for 24 h. See also SI Appendix , Fig. S6 A . ( B ) Mean ± SEM percentage of cells with overlapping calnexin and IgM signals as in Fig. 6 D and E of P3HR-1 cells expressing the indicated control or HRD1 sgRNA, uninduced or 4-HT induced in the absence or presence of bortezomib as indicated, using data from 15 randomly selected panels of 300 cells from n = 3 replicates. Cells with IgM-calnexin colocalization were defined by those with colocalization signal above the threshold defined in mock-induced control cells. Images were analyzed using the ImageJ ComDet plugin. ( C ) Analysis of BALF0/1 driven IgM HC ubiquitination. Immunoblot analysis of 1% input and anti-EGFP immunopurified complexes from 293T cells transiently expressing EGFP-IgM HC (EGFP–IgM–HC) together with BALF0/1 or BXLF1 and blotted for poly-ubiquitin chains, EGFP, or HA, as indicated. ( D ) Analysis of BALF0/1-mediated BCR degradation dependency on E1 ubiquitin ligase activity. Immunoblot analysis of WCL from P3HR-1 cells 4-HT induced into lytic cycle for 24 h, in the absence or presence of increasing doses of the ubiquitin E1 enzyme inhibitor TAK-243 (1 and 2 µM). See also SI Appendix , Fig. S7 F . Statistical analysis was performed with Student’s t test unless otherwise specified. ** P < 0.01. Blots in A , C , and D are representative of ≥n = 2 independent experiments. See also SI Appendix , Figs. S4–S7 .

To examine ubiquitin–proteasome pathway roles in EBV lytic cycle BCR depletion, we tested whether BALF0/1 expression was sufficient to induce polyubiquitination of Ig HC complexes. BALF0/1 or control BXLF1 cDNAs were expressed in human embryonic kidney (HEK) 293 cells, together with an EGFP-tagged IgM HC construct. High molecular weight poly-ubiquitin conjugates were evident upon immunoblotting of immunopurified EGFP–IgM–HC complexes from BALF0/1, but not from control BXLF1 expressing cells ( Fig. 4 C ). Furthermore, small-molecule TAK-243 ( 48 ) inhibition of the ubiquitin activating enzyme E1 ligase, which is the first enzyme in most host cell ubiquitin cascades, also significantly impaired BCR degradation in lytic-induced P3HR-1 cells ( Fig. 4 D and SI Appendix , Fig. S7 F ). These data suggest that BALF0/1 subverts ERAD to route BCR to cytoplasmic proteasomes ( SI Appendix , Fig. S7 G ).

BALF0/1 BCR Internalization and Degradation Requires the Ig HC Cytoplasmic Tail.

We next characterized how BALF0/1 recognizes BCR complexes. Notably, CD79A and CD79B contain 61- and 48-amino acid cytoplasmic tails, respectively, whereas Ig HCs have as few as three cytoplasmic tail residues ( Fig. 5 A ). We therefore hypothesized that BALF0/1 recognizes CD79A and/or CD79B cytoplasmic tails. To test this, we expressed sgRNA against CD79A or CD79B in P3HR-1 cells and then measured IgM abundance upon lytic reactivation. Despite substantial CD79A or CD79B depletion, EBV lytic induction targeted IgM for degradation ( SI Appendix , Fig. S8 A – D ). However, as BALF0/1 may redundantly use either CD79A or CD79B to target BCR, we next depleted both CD79A and CD79B in P3HR-1. EBV nonetheless targeted IgM for degradation at comparable efficiency as in control cells ( Fig. 5 B and SI Appendix , Fig. S8 E ), as did BALF0/1 expression ( Fig. 5 C and SI Appendix , Fig. S8 F ). Similar results were observed by confocal microscopy ( SI Appendix , Fig. S8 G and H ). We observed in a similar degree of IgM/calnexin colocalization in 4-HT/bortezomib-treated control and CD79A/CD79B double KO cells ( SI Appendix , Fig. S9 A – E ). BALF0/1 also targeted IgM for degradation in transiently transfected 293T cells, in the absence of CD79A/B ( Fig. 5 D ). Even though IgM is retained in the ER in the absence of CD79A/B ( 49 – 51 ), this result provides evidence that BALF0/1 can target ER-resident IgM for degradation. Taken together, these results indicate that CD79A/B are dispensable for BALF0/1-mediated degradation of membrane-bound immunoglobulin, much of which resides at the PM in latently infected Burkitt cells.

BALF0/1-mediated BCR degradation requires the Ig HC cytoplasmic tail. ( A ) Schematic model of the BCR complex immunoglobulin HC (IgH), LC, and the CD79A and CD79B signaling chains. The Inset shows the IgM cytoplasmic tail residues. ITAM, immunoreceptor tyrosine–based activation motif. ( B ) Analysis of CD79A/B role in BALF0/1-mediated BCR targeting. Immunoblot analysis of WCL from Cas9+ P3HR-1 cells expressing the indicated CD79A and CD79B sgRNA, 4-HT induced into lytic cycle for 24 h. See also SI Appendix , Fig. S8 E . ( C ) Analysis of whether CD79A/B are necessary for BALF0/1-mediated BCR degradation. Immunoblot analysis of WCL from Cas9+ P3HR-1 cells expressing the indicated CD79A and CD79B sgRNA and BALF0/1 or BXLF1 cDNA. See also SI Appendix , Fig. S8 F . ( D ) Analysis of the IgM cytoplasmic tail KVK motif role in BALF0/1 targeting. Immunoblot analysis of WCL from 293T cells transiently expressing wildtype (WT) or cytoplasmic tail deleted (∆KVK) EGFP-tagged IgM HC (IgM-HC), alone or together with either BXLF1 or BALF0/1. See also SI Appendix , Fig. S9 F . Blots in B – D are representative of ≥n = 2 replicates. See also SI Appendix , Figs. S8 and S9 .

We hypothesized that the Ig HC cytoplasmic tail is instead critical for BALF0/1-mediated BCR degradation. Notably, the IgM (mu) and IgG (gamma) HCs contain only 3- and 28 residue cytoplasmic tails, respectively, including a conserved KVK motif, which has roles in immunoglobulin association with the cytoskeleton and internalization ( 52 – 55 ) ( Fig. 5 A ). To test whether the IgM cytoplasmic tail was necessary for BALF0/1 targeting, we transiently expressed either WT or C-terminal tail deletion mutant (∆KVK) IgM HCs in 293T cells, together with either BALF0/1 or BXLF1. Immunoblot analysis showed a clear reduction of WT, but not ∆KVK IgM HC abundance in cells coexpressing BALF0/1, but not BXLF1 ( Fig. 5 D and SI Appendix , Fig. S9 F ). The IgA and IgE cytoplasmic tails do not contain a KVK motif, and IgA HC levels were not diminished by coexpression with BALF0/1 ( SI Appendix , Fig. S9 G and H ). Importantly, addition of KVK residues to the IgA cytoplasmic tail was sufficient to sensitize IgA for BALF0/1 triggered degradation ( SI Appendix , Fig. S9 I and J ). Thus, BALF0/1 may target one or both KVK lysine residues for ubiquitination to initiate BCR degradation, likely together with a host ubiquitin ligase.

BALF0/1 Supports Virions Release from Cells Expressing an Anti-gp350 Antibody.

Tegumented EBV capsids acquire their lipid envelope by budding into the Golgi or Golgi-derived secretory pathway vesicles ( 56 ) prior to secretion from the cell ( 57 ). Therefore, maturing EBV virions travel through the same secretory pathway compartments as immunoglobulin, and in the absence of degradation, membrane-bound immunoglobulin is incorporated into the virion envelope. We hypothesized that BALF0/1 may have evolved to target the immunoglobulin, in order to cope with the situation where EBV infects a B cell with BCR reactive to EBV glycoproteins, which could prevent virions release via tethering particles to the PM or could induce clumping.

To test this, we stably expressed a membrane-bound IgG isoform of the murine monoclonal antibody 72A1, which reacts with EBV gp350, in EBV+ Akata cells ( 58 ). We used a humanized 72A1 HC, in which the murine constant region was replaced by the human Fc residues ( 58 ). We established BXLF1 vs. BALF0/1 edited 72A1+ Akata cells, which expressed 72A1 at a somewhat lower level than endogenous IgG ( SI Appendix , Fig. S10 A ), and triggered them for EBV lytic reactivation by Ig cross-linking. We then measured relative EBV intracellular and extracellular genome copy number at 24 h poststimulation, an early timepoint prior to the onset of cell death, which could otherwise nonspecifically release virions. As expected, FACS analysis demonstrated significantly lower levels of PM 72A1 in BXLF1 KO than in BALF0/1 KO cells, as detected by antibody reactive with murine LC and HC ( Fig. 6 A ). Interestingly, at 24 h post Ig-crosslinking, we observed increased intracellular and decreased extracellular EBV genome copy number in BALF0/1 KO cells, relative to levels in BXLF1 KO controls ( Fig. 6 B and C ). Similarly, BALF0/1 KO resulted in a threefold decrease in infectious virions released from Akata cells relative to BXLF1 KO control cell levels, as measured by coculture of the Ig-crosslinked Akata supernatants with Daudi target cells [the green Daudi assay ( 59 )] ( SI Appendix , Fig. S10 B ). It is worth noting that BALF0/1 KO did not affect late gene gp350 expression, suggesting an intact lytic cycle ( SI Appendix , Fig. S10 C ), and likewise did not affect cell viability ( SI Appendix , Fig. S10 D ), indicating that differences in extracellular viral load were not based on differences in live cell number. We next expressed a control monoclonal IgM antibody against influenza HA in BXLF1 vs. BALF0/1 edited EBV+ Akata cells ( SI Appendix , Fig. S10 E ). EBV reactivation reduced PM IgM levels to a greater degree in control BXLF1 KO than in BALF0/1 KO Akata cells, as expected ( SI Appendix , Fig. S10 F ). Levels of encapsidated, extracellular EBV genomes did not significantly differ between BXLF1 vs. BALF0/1 edited cells ( SI Appendix , Fig. S10 G ).

BALF0/1 supports virions release from cells expressing EBV gp350-specific antibody. ( A ) Analysis of monoclonal antibody 72A1 targeting by the EBV lytic cycle in BXLF1 vs. BALF0/1 edited cells. Mean ± SEM of PM 72A1 antibody abundance from n = 3 replicates of Cas9+ Akata cells expressing the indicated sgRNA and stably expressing membrane-bound anti-gp350 monoclonal IgG antibody 72A1. The human-murine chimera 72A1 was detected by Alexa Fluor™ 488 tagged goat anti-mouse IgG secondary antibody. ( B ) Analysis of 72A1 effects on intracellular EBV genome copy # in BXLF1 vs. BALF0/1 edited cells. qRT-PCR analysis of EBV intracellular genome copy number from EBV+ Cas9+ Akata cells with stable membrane-bound 72A1 antibody expression together with BXLF1 or BALF0/1 sgRNA, induced into lytic replication by αIgG-crosslinking for 24 h. ( C ) Analysis of 72A1 effects on extracellular EBV genome copy # in BXLF1 vs. BALF0/1 edited cells. qRT-PCR analysis of EBV extracellular genome copy number from EBV+ Cas9+ Akata cells with stable membrane-bound 72A1 antibody expression together with BXLF1 or BALF0/1 sgRNA, induced into lytic replication by αIgG-crosslinking for 24 h. ( D ) Analysis of BALF0/1 effects on BCR-independent P3HR-1 vs. BCR-dependent HBL1 survival. Mean ± SEM percentages from n = 3 replicates of live P3HR1 or HBL1 cells mock induced or induced for BALF0/1 for 24 h, as determined by trypan blue exclusion assay. Statistical analysis was performed with Student’s t test unless otherwise specified. **** P < 0.0001, *** P < 0.001, ** P < 0.01, * P < 0.05, ns non-significant . See also SI Appendix , Fig. S10 .

To test the hypothesis that BALF0/1 prevents incorporation of immunoglobulin into virion envelopes, we induced BXLF1 vs. BALF0/1 edited Akata cells into lytic replication by Ig-crosslinking for 24 h. We then immunopurified virus particles from cell culture supernatants via pulldown with an antibody specific for the EBV glycoprotein gp42. Immunoblot analysis of purified EBV revealed incorporation of human IgG into virus particles produced in BALF0/1 edited Akata, but to a lesser extent in BXLF1 edited Akata ( SI Appendix , Fig. S10 H ). The integral PM proteins CD19 and CD21 were present at similar levels in EBV virions produced from BXLF1 vs. BALF0/1 edited cells, whereas the soluble CD79A or CD79B BCR components were not detected ( SI Appendix , Fig. S10 H ).

BALF0/1 Impairs Survival of Non-Hodgkin Lymphoma Dependent on BCR Signaling.

Several subtypes of non-Hodgkin lymphoma, including activated B cell–like (ABC) and germinal center B cell (GCB) subtypes of diffuse large B cell lymphoma (DLBCL), mantle cell lymphoma, and Burkitt lymphoma subtypes depend on constitutive BCR signaling for survival ( 60 – 65 ). We therefore hypothesized that BALF0/1 expression would be sufficient to trigger cell death of BCR-dependent B cells. We inducibly expressed BALF0/1 in HBL1 diffuse large B cell lymphoma cells, which are dependent on BCR signaling for survival ( 66 ). As a control, we expressed BALF0/1 in P3HR-1, which we did not find to be dependent on BCR components in a CRISPR screen ( 67 ). While BCR abundance was similarly reduced by BALF0/1 expression in P3HR-1 and in HBL1 ( SI Appendix , Fig. S10 I – L ), BALF0/1 significantly reduced HBL1 but not P3HR-1 live cell number ( Fig. 6 D ). Collectively, our studies support a model in which BALF0/1 triggers BCR complexes for degradation ( Fig. 7 ).

Schematic model of BALF0/1-mediated BCR degradation. BALF0/1 recognizes IgM or IgG HC cytoplasmic tails in a manner dependent on their KVK motif, induces caveolin-1-dependent BCR endocytosis and delivery into caveosomes. BCR are then retrograde trafficked to the ER and dislocated to the cytosol, where they are degraded by proteasomes. In the absence of BALF0/1, EBV release is impeded from B cells with BCR reactive with virion glycoproteins. Therefore, delivery of cell-permeable BALF0/1 to B cell lymphomas dependent on BCR signaling could represent a novel therapeutic approach.

Discussion

As a B cell tropic herpesvirus, EBV most cope with large quantities of immunoglobulin as it replicates within and egresses from host B cells. Here, we identified that the EBV early lytic protein BALF0/1 targets the BCR complex for degradation. Rather than routing membrane-bound BCR to lysosomes, BALF0/1 instead subverts the caveolin-dependent endocytosis and ERAD pathways to target BCR for ubiquitin-dependent proteasomal degradation. We hypothesize that EBV evolved this strategy in order to facilitate release of virions from cells that express BCR reactive to an EBV virion component, which would otherwise impair virus assembly and/or release, for instance by tethering or causing clumping of egressing virus. In this manner, BALF0/1 may act analogously with HIV Nef and Vpu, which downmodulate CD4. This is necessary to support HIV release from CD4+ T cells, since HIV utilizes CD4 as a major entry receptor, but must avoid being tethered by it as it egresses from CD4+ T cells ( 68 – 70 ). It would also be similar to viral evasion of the host restriction factor tetherin, which tethers viral particles to host cell PM to impair their release ( 71 ).

Why then does EBV not target CD21 (CR2) for degradation, since CD21 association with gp350 could also inhibit EBV release? It is plausible that gp350 is produced in excess of CD21, such that CD21 is frequently bound by free gp350 rather than virion-incorporated gp350. Alternatively, it is possible that CR2 is expressed at low levels in the EBV+ memory cell population that undergoes reactivation. For instance, while CD21 is dynamically expressed in B cells, typically in common with surface immunoglobulin expression, CD21 is expressed at low levels in specific subsets of disease-associated memory cells ( 72 – 74 ). It will therefore be of interest to define levels of BCR subunits including CD21 in the rare population of EBV+ memory cells in vivo. Furthermore, the affinity between anti-EBV glycoprotein antibodies and their EBV glycoprotein targets may be significantly higher than that between CR2 and gp350. Similarly, gp42 levels are lower in virions secreted from MHC Class II expressing B cells than from epithelial cells, as endogenously synthesized class II molecules can trap gp42 before it is incorporated into the viral envelope ( 75 ).

The observation that BALF0/1 targets BCR raises the question of how frequently EBV may infect B cells reactive to viral glycoproteins. Germline encoded antibodies reactive to public viral epitopes, including multiple reactive with EBV gp350, were recently described ( 76 ). While EBV may encounter such B cells at low frequency at the outset of primary infection, antibodies against virion structural proteins including gp350 and viral capsid antigen (VCA) are readily detected in EBV-infected individuals, suggesting the expansion of a pool of B cells reactive with EBV antigens. In fact, anti-VCA antibodies are among the earliest detected upon primary EBV infection and are used to establish the diagnosis of mononucleosis. Anti-gp350 and VCA antibodies can also be abundant in patients with EBV-associated cancers ( 77 – 82 ). As acute EBV infection progresses over the course of many months, it may be increasingly common for EBV to encounter a target B cell with BCR reactive to viral components. For instance, we note that as many as 20% of tonsillar T cells can be reactive with EBV-encoded peptides in hosts with infectious mononucleosis ( 83 ), suggesting that high rates of EBV-reactive B cells may be likely also be present. Furthermore, BCR reactive to EBV virion components may facilitate viral adhesion to and/or entry, potentially serving as coreceptors. B cells invest considerable resources in immunoglobulin production. Another potential advantage of EBV lytic cycle degradation of high abundance BCR is that it may provide amino acid building blocks to support viral replication. It is also possible that EBV targets BCR to block tonic signaling upon lytic reactivation.

BCR are internalized via clathrin-mediated endocytosis upon antigen binding ( 84 – 86 ), which delivers engulfed antigens to the lysosome for MHC Class II loading. EBV BALF0/1 instead triggers BCR internalization through a caveolar pathway reminiscent of the cholera toxin entry pathway ( 39 ). The principal structural component of caveolae, caveolin-1, is regulated by Src family kinase phosphorylation ( 37 , 87 – 91 ). Interestingly, our recent proteomic analysis identified that BALF0/1 associates with the serine/threonine protein phosphatase 2A (PP2A) regulatory subunit PP2R1A and also with the protein phosphatase 6 (PP6) catalytic subunit PPP6C ( 22 ). Since PP2A dephosphorylates Src kinases ( 92 ), and since the phosphatase inhibitor okadaic acid stimulates caveolae mobilization ( 93 – 96 ), we speculate that BALF0/1 may subvert host phosphatase as well as ubiquitin pathways to trigger BCR internalization. Calnexin associates with and participates in the folding of newly synthesized membrane-bound Ig within the ER ( 97 , 98 ), and the calnexin membrane anchoring was important for association with IgM ( 99 ). Furthermore, glucose trimming and reglucosylation cycles can regulate association of ER-resident protein cargo with calnexin ( 100 ), and glucose and mannose trimming are implicated in MHC class I HC turnover by ERAD ( 101 ).

Multiple lines of evidence implicate the ERAD pathway downstream of BALF0/1, including demonstration that HRD1, p97/VCP, and E1 ubiquitin ligase activity are important for immunoglobulin turnover. While we did not obtain evidence that BAP31 is critical for BALF0/1-mediated BCR degradation, we speculate that it may nonetheless play an important role in particular contexts. It is plausible that another host factor is redundant with BAP31 in Burkitt cells, but perhaps not in memory B cells undergoing plasma cell differentiation, a physiological site of EBV reactivation for which models are unavailable. BAP31 copurifies with B cell receptors ( 102 ), and multiple viral proteins associate with BAP31, including Kaposi’s sarcoma–associated herpesvirus K3 and K5 ( 102 , 103 ). We note that not all ERAD pathways are dependent on BAP31. BALF0/1 triggers BCR degradation via a similar pathway to that subverted by cholera toxin, which following caveolar uptake, travels retrograde to the ER, from where it is routed to the cytosol ( 39 ). Our data indicate that BALF0/1 stimulate ubiquitination of the IgM HC cytoplasmic tail, likely at one or both KVK motif lysines, and this is necessary for BCR degradation, since the E1 Ub ligase inhibitor TAK-243 rescued BCR abundance in the EBV lytic cycle. Since BALF0/1 is not known to have ubiquitin ligase activity, it instead likely recruits an E3 ligase. Further studies are required to define whether Ig HC ubiquitination is required for retrograde trafficking to the ER.

Multiple human lymphomas are dependent on BCR signaling, including MCD DLBCL, chronic lymphomatous leukemia, and subtypes of mantle cell and Burkitt lymphoma ( 60 – 65 ). Consequently, the BCR pathway Bruton’s tyrosine kinase is a key B cell lymphoma therapeutic target, including by the small-molecule inhibitor ibrutinib ( 104 ). Since BCR knockdown is toxic to BCR-dependent lymphomas ( 60 , 61 ), our results suggest that BALF0/1, fused to cell-penetrating peptides could be a therapeutic approach. With regard to limitations of our study, primary B cell EBV lytic reactivation models are unavailable, including of memory or plasma cells. Our studies were therefore limited to Burkitt B cell models. Similarly, while multiple experiments were performed in the context of overexpression of EBV immediate early protein Z-HT and R-HT alleles, key phenotypes were confirmed with lytic induction by immunoglobulin cross-linking. Deletion of the KVK motif may have introduced additional effects on IgM beyond its interaction with BALF0/1 that for example may have altered its trafficking, and yet we were able to demonstrate that addition of KVK sensitized IgA to BALF0/1-mediated degradation.

In summary, we used a recently constructed EBV lytic cycle protein–protein interaction map to identify that the EBV early lytic protein BALF0/1 targets the BCR complex for degradation. BALF0 and BALF1 each exhibited this property and subverted the caveolin-1 pathway to internalize BCR complexes, which were trafficked in a retrograde manner to the ER, where they underwent proteasomal degradation via the ERAD pathway in a p97/VCP and ubiquitin-dependent manner. The IgM cytoplasmic tail KVK motif, present also in the IgG cytoplasmic tail, was necessary for BCR degradation, enabling BALF0/1 to selectively target IgM and IgG.

Methods and Materials

See SI Appendix for additional details.

Chemicals and Compounds.

Unless otherwise specified, P3HR-1 Z-HT/R-HT or Akata cells lytic induction was triggered by addition of 400 nM 4-hydroxy tamoxifen (4-HT) (Sigma #68392-35-8) for 24 h or by 15 µg/mL α-human IgG (Agilent #A042301-2) for 48 h, respectively. 5 µg/mL doxycycline were used to induce BALF0/1 and BARF1 cDNA expression for 24 h. All inhibitors were used where indicated and were added simultaneously with the lytic or cDNA stimulus. Details in SI Appendix .

cDNA Cloning.

Unless otherwise specified, all cDNA were cloned into expression vectors by Gateway recombination. Briefly, 150 ng of the donor vector containing the EBV cDNA and the destination vector were coincubated with 1X LR Clonase Enzyme Mix (Invitrogen #11789-020) overnight at room temperature. The reaction mixture was then transformed into 50 μL of Stbl3 Escherichia coli , spread on LB plates with ampicillin. Entry vectors containing BALF0/1 and BXLF1 cDNA were gifts from Eric Johannsen. Entry vectors with IGHM and IGHA1 cDNA were obtained from DNASU (IGHM # HsCD00514313; IGHA1 # HsCD00512453 respectively). cDNA for expressing the anti-gp350 monoclonal antibody 72A1 HC and LC were gifts from Elliott Kieff and Fred Wang ( 58 ). Plasmids pDM137 and pDM066, which encode anti-influenza hemagglutinin (HA) IgM HC and LC, were gifts from Daniel Maurer and Aaron Schmidt. pDM137 and pDM066 contains the CH67 HC that expresses human membrane-bound IgM and LC, respectively. Details for cloning in SI Appendix .

Immunoblot and Immunoprecipitation.

Immunoblot analyses were performed as described previously ( 28 ). In brief, whole cell lysates were separated by SDS-PAGE, transferred onto nitrocellulose membranes, blocked with 5% nonfat dry milk in TBST buffer for 1 h and incubated with primary antibodies (1:1,000) at 4 °C overnight. Blots were washed three times in TBST and incubated with secondary antibodies for 1 h at room temperature. Blots were then washed three times in TBST solution, incubated with ECL chemiluminescence solution (Thermo Fisher #34578) and images were captured by Licor Fc platform. For poly-ubiquitinylation analysis, 5 nM bortezomib was added to the cells 6 h posttransfection of 293T cells. After 24 h, cells were then lysed in ice-cold lysis buffer (1% v/v NP40, 150 mM Tris, 300 mM NaCl) supplemented with 1X cOmplete™ EDTA-free protease inhibitor cocktail, 1 mM Na 3 VO 4 , 1 mM NaF, 1 mM PMSF, 4 mM 1, 10 o-phenanthroline, 2 mM sodium pyrophosphate and 1 mM EDTA for 1 h at 4 °C with rotation. Lysed cells were precleared with 2 μg/mL anti-rabbit isotype IgG antibody (Cell Signaling #2729) with protein A/G magnetic beads (Pierce, Thermo) for 1 h at 4 °C with rotation, which was then incubated with 2 μg/mL anti-GFP antibody (Proteintech #50430-2-AP) for 1 h at 4 °C with rotation. Protein A/G magnetic beads were then added to the immunocomplex and were coincubated at 4 °C overnight with rotation. Beads were washed with lysis buffer four times and were eluted using 1X SDS loading buffer incubated for 10 min at 95 °C. Details in SI Appendix .

Analysis of Human IgG in EBV Virions.

Supernatant containing EBV virions were collected from uninduced or lytic-induced (72 h) Akata-EBV-Cas9+ cells that express sgBXLF1 or sgBALF0/1. They were passed through a 0.45 µm filter and precleared with 2 μg/mL anti-rabbit isotype IgG antibody (Cell Signaling #2729) with protein A/G magnetic beads (Pierce, Thermo) for 1 h at 4 °C with rotation. EBV virions immunoprecipitation was performed by incubating the precleared supernatant with 2 μg/mL anti-EBV BZLF2 (gp42) rabbit polyclonal antibody (Thermo Fisher # PA5-117635) for 4 h at 4 °C with rotation. Protein A/G magnetic beads were then added to the immunocomplex and were coincubated at 4 °C overnight with rotation. Beads were washed with lysis buffer four times and were eluted using 1X SDS loading buffer incubated for 10 min at 95 °C. Immunoblot analysis was performed as stated above.

Flow Cytometry Analysis.

Cells were washed once with FACS buffer (PBS supplemented with 2% fetal bovine serum). Cells were then incubated with the corresponding antibody at 1:500 for 30 min at 4 °C. Cells were pelleted, washed twice, and resuspended in FACS buffer. For 7-AAD assays (Thermo Fisher, Cat#A1310), cells were harvested and washed twice with FACS buffer. Cells were incubated with a 1 μg/mL 7-AAD solution for 5 min at room temperature, protected from light. Cells were then analyzed via FACS on a BD FACSCalibur instrument and analyzed by FlowJo V10.

Immunofluorescence Analysis.

Burkitt lymphoma cells dried on glass slides were fixed with 4% paraformaldehyde/PBS solution for 10 min, permeabilized with 0.5% Triton X-100/PBS for 5 min and blocked with 1% BSA/PBS for 1 h at room temperature. Cells were then incubated with a cocktail of primary antibodies (1:100 dilution) in blocking solution for 1 h at 37 °C. Cells were then washed twice with PBS and incubated with a cocktail of secondary antibodies at 1:1,000 in PBS for 1 h at 37 °C in the dark. Finally, cells were washed twice with PBS and were stained with DAPI and mounted overnight with ProLong™ Gold Antifade. Image acquisition and analysis was performed with a Zeiss LSM 800 instrument and with Zeiss Zen Lite (Blue) software, respectively.

PM staining was performed using MemBrite Fix Cell Surface Staining Kits (Catalog#: 30098-T, Biotium), according to the manufacturer’s protocol. After removing culture medium, cells were resuspended with 1X Pre-Staining Solution buffer for 10 min at 37 °C. Cells were then pelleted and stained with 1X Stain Solution buffer for another 10 min at 37 °C. After two washes with PBS, cells were fixed with 4% paraformaldehyde in PBS for 20 min, followed by permeabilization with 0.1% Triton X-100 in PBS for 10 min at room temperature. Then, immunofluorescence staining was performed as stated above. Details for quantification in SI Appendix .

EBV Genome Copy Number Quantification.

For intracellular EBV DNA extraction, total DNA from 1 × 10 6 cells was extracted using the Blood & Cell Culture DNA Mini Kit (Qiagen). Extracted DNA was diluted to 10 ng/μL and subjected to qPCR analysis, targeting the EBV BALF5 gene. For extracellular viral DNA extraction, supernatants from Akata EBV+ cells induced for lytic replication were collected and incubated with 25 mg/mL of DNase for 15 min at 50 °C, to degrade nonencapsidated DNA. Viral DNA was extracted by the Blood & Cell Culture DNA Mini Kit (Qiagen) and samples were subjected to qPCR analysis. Details in SI Appendix .

Quantification and Statistical Analysis.

Unless otherwise indicated, all bar graphs and line graphs represent the arithmetic mean of three independent experiments (n = 3), with error bars denoting SEM. Significance between the control and experimental groups, or indicated pairs of groups, was assessed using the unpaired Student’s t test in the GraphPad Prism 7 software. P values correlate with symbols as follows, unless otherwise indicated: ns = not significant, P > 0.05; * P < 0.05; ** P < 0.01; *** P < 0.001; **** P < 0.0001. Pathway analysis was performed and visualized by using DAVID. Cytoscape version 3.8.1 was used to construct protein–protein interaction networks.

Schematic Models.

Biorender was used to create all schematic models.

Supplementary Material

Appendix 01 (PDF)