R9AP is a functional receptor for Epstein-Barr virus infection in both epithelial cells and B cells

Epstein-Barr virus (EBV), also known as the rst human tumor virus, is linked to about 200,000 new cancer cases and millions of non-malignant diseases every year. EBV infects both human epithelial cells and B cells. Several virally encoded glycoproteins dene tropism and mediate a complicated entry process. Here, we show that in both epithelial cells and B cells, R9AP silencing or genetic knockout signicantly inhibits EBV infection, whereas R9AP overexpression promotes EBV infection, establishing R9AP as an essential entry receptor for EBV. Mechanistically, R9AP directly binds to EBV glycoproteins gH/gL to mediate membrane fusion. Importantly, the interaction of R9AP with gH/gL is inhibited by the highly potent, competitive gH/gL neutralizing antibody AMMO1 that blocks EBV infection of both epithelial cells and B cells. Furthermore, a R9AP peptide encompassing the gH/gL binding site inhibits EBV infection in vitro and reduces viral load in EBV infected humanized mice. Altogether, we propose R9AP as the rst characterized receptor for EBV infection common to epithelial cells and B cells and a potential target for intervention. Immunoprecipitation and GST pull-down assay. For the co-immunoprecipitation assay, HEK-293T cells were transfected with the indicated plasmid and lysed in lysis containing 1%NP-40 (N885726; 150mM NaCl (13423-6X1KG-R; Invitrogen, California), 2.5mM EDTA (E6758; Sigma-Aldrich, Germany), 20mM HEPES (H7523; Sigma-Aldrich, Germany) pH7.4 and protease inhibitor cocktail at 36h post transfection. The lysates were cleared by centrifugation at 12000 rpm and 4°C 10min . The supernatants were incubated with ANTI-FLAG M2 Gel (A2220, Sigma-Aldrich, Germany) or Anti-c-Myc Agarose Anity Gel (A7470, Sigma-Aldrich, Germany) overnight. Then, the gels with bound protein were washed three times with lysis buffer and subjected to WB analysis. For the GST pull-down assay, GST-R9AP 1-210 and His-gH/gL were incubated in lysis buffer overnight, then washed three times with lysis buffer and analyzed by WB. For the antibody competition binding assay, HEK-293T cells were transfected with plasmids encoding FLAG-R9AP or Myc-gH/gL for 36h, and lysed in lysis buffer. HEK-293T cells transfected with empty vector were used as control. Lysates containing Myc-gH/gL protein were incubated with 5mg IgG control, 5mg Ammo1, 10mg Ammo1, 5mg CL59 or 10mg CL59 overnight, and then incubated with lysates containing FLAG-R9AP or the control overnight. Finally, Myc-gH/gL was pulled down using Anti-c-Myc Agarose Anity Gel which was then washed three times with lysis buffer and subjected to WB analysis. in vitro experiments and detection of EBV DNA copy numbers in mice at the 6 th week (ns, not signicant; *P< 0.05, **P< 0.01, ***P< 0.001). A Log-rank test was used for statistical analysis involving two group comparisons for survival in the in vivo experiments. Statistical analyses were performed using GraphPad Prism (GraphPad Software, San Diego, CA).


Introduction
Epstein-Barr virus (EBV) is present in more than 90% of the global human population. In adolescents and young adults, primary EBV infection frequently caused infectious mononucleosis (IM) [1][2][3] . EBV is also involved in the development of autoimmune diseases, such as multiple sclerosis (MS). MS is the most common autoimmune disorder affecting the central nervous system [4][5][6] . Moreover, as the rst virus discovered to cause human cancer, EBV is associated to 200,000 new cancer cases annually, including B cell-derived malignancies, such as Burkitt lymphoma and Hodgkin lymphoma, as well as epithelial tumors, such as nasopharyngeal carcinoma (NPC) and EBV-positive gastric carcinoma [7][8][9] . Despite numerous breakthrough studies, there is still no licensed vaccine that can prevent primary EBV infection 10 .
It has been suggested that EBV uses different combinations of viral glycoproteins to infect human epithelial cells and B cells 11 . The attachment of EBV to B cells depends on the interaction between EBV glycoprotein gp350 and B cell complement receptor CR2 (also known as CD21) or CD35 12,13 . Recently, CD21 also has been reported as the receptor interacting with EBV in T cells 14 . It has been proposed that EBV glycoprotein gp42, in complex with gH/gL, interacts with HLA class II to trigger the fusion activity of gB in B cells [15][16][17] .
In contrast to B cells, epithelial cells lack not only CD21 and CD35 but also HLA class II 18,19 , suggesting that EBV relies on different mechanisms and glycoproteins to infect epithelial cells [20][21][22] . The attachment of EBV to epithelial cells may involve EBV gH/gL or BMRF2 and their fusion may involve gH/gL and gB 11 . The interaction of EBV BMRF2 with integrins mediates virus attachment and entry to polarized epithelial cells 23 . EBV gH/gL also has been found to interact with integrins to trigger fusion 24,25 . We previously showed that gH/gL interacts with nonmuscle myosin heavy chain IIA (NMHC-IIA) to mediate EBV attachment to B lymphoma Mo-MLV insertion region 1 homolog (Bmi1)-immortalized nasopharyngeal epithelial cells (NPECs) that grow as sphere-like cells (SLCs) instead of monolayer cells (MLCs) 26,27 . Currently, there is no evidence supporting a role of integrins or NMHC-IIA in EBV infection of B cells. Our previous work showed that neuropilin 1 (NRP1) could directly interact with EBV gB to promote EBV infection of NPECs but not B cells, which lack NRP1 expression 28 . Chen and we also identi ed ephrin receptor A2 (EphA2) bound to gH/gL and gB to serve as a receptor for epithelial cells but not B cells 29,30 . It is unknown if there are additional receptors that mediate EBV infection of epithelial cells.
As members of the "core fusion machinery" 31 , gH/gL and gB are required for EBV to infect B cells and epithelial cells. Functional regions in EBV gH can be divided into four domains: domain (D)-I, D-II, D-III, and D-IV 32 . gL is important for gH folding and intertwines with gH D-I 33 . Several monoclonal antibodies have been identi ed to bind with different domains of gH to block EBV infection in either B cells or epithelial cells. However, AMMO1 is so far the only reported monoclonal antibody that binds gH/gL and blocks not only EBV infection of both epithelial cells and B cells in vitro but also EBV infection in animal infection models 34,35 . AMMO1 interacts with a discontinuous epitope among gH/gL D-I and D-II, but its precise mechanism of neutralization is unknown. It is possible that AMMO1 blocks the interaction of an unidenti ed common cellular factor with gH/gL. To identify the potential common cellular factor, we used cDNA microarray followed by siRNA screening and identi ed R9AP as a common receptor for EBV infection in both epithelial cells and B cells.

R9AP promotes EBV infection
We reported that EBV infection e ciency was much higher in Bmi1-immortalized nasopharyngeal epithelial cells (NPECs) grown as sphere-like cells (SLCs) than as monolayer cells (MLCs) 27 . We speculated that some of the key factors that mediated EBV infection might be upregulated in SLCs than in MLCs. Thus, using microarray analysis we identi ed 132 genes that were upregulated in SLCs from two Bmi1-immortalized NPEC cell lines, namely NPEC1-Bmi1 and NPEC2-Bmi1 (Extended Data g. 1, Supplementary Table 1). To identify the potential receptors or co-receptors associated with EBV infection, we performed a siRNA screening in SLCs targeting the 72 genes encoding membrane-associated proteins among these 132 genes with a pool of 4 siRNAs targeting each gene. The 72 siRNA pools were rst transfected into NPEC2-Bmi1 MLCs, which were then grown as SLCs and infected with a recombinant EBV that expressed eGFP (EBfaV-GFP) 36 . EBV positive cells were identi ed by ow cytometry. By comparing the percentage of infected cells among the siRNA transfected cells with that of the control siRNA transfected cells, 10 genes were identi ed with a more than 50 percent reduction of EBV infection (Fig. 1a). These 10 siRNAs pools were then rescreened in NPEC1-Bmi1 cells for their effect on EBV infection. Targeting CNGA1, GPR1, SLC26A9 and R9AP were con rmed to reduce EBV infection more than 50 percent (Fig. 1b). RT-qPCR assay showed that each of these siRNAs pools e ciently depleted targeted genes expression (Extended Data g. 2). We next examined the expression of these genes in several cell lines that are susceptible to EBV infection. Western blotting showed that human B-cell lines, including EBV-positive Akata, EBV-negative Akata, and Raji, nasopharyngeal epithelial cell lines, including NPEC1-Bmi1, HNE1 and HK1, and the gastric epithelial cell line AGS consistently expressed CNGA1 and R9AP, but not GPR1 or SLC26A9 (Fig. 1c). To test whether overexpression of CNGA1 or R9AP could promote EBV infection, we transfected human embryonic kidney 293T (HEK-293T) cells with a construct expressing CNGA1 or R9AP (Fig. 1d). We found that overexpression of R9AP increased EBV infection more than 3 folds (Fig. 1e), we therefore selected R9AP for further investigation. A diagram summarizing the screening process is presented in Fig. 1f.
R9AP interacts with EBV gH/gL but not gB EBV gH/gL and gB are considered to be necessary for EBV infection of both epithelial cells and B cells. To examine the potential interaction of R9AP with EBV envelope glycoproteins, co-immunoprecipitations were performed using HEK-293T cells transfected with a vector expressing FLAG-R9AP together with a vector expressing Myc-gH plus gL, Myc-gH alone, or Myc-gB or with an empty vector. Immunoprecipitation with anti-Myc antibody identi ed an association between gH/gL and R9AP but not gB. The interaction was strikingly reduced without the expression of gL (Fig. 2a), consistent with the notion that gL is essential for assisting the folding and transportation of gH 11 . Consistently, reverse immunoprecipitation with anti-FLAG antibody con rmed the interaction between gH/gL and R9AP (Fig. 2b). Puri ed recombinant His-gH/gL was pulled down by puri ed recombinant GST-R9AP 1-210 , demonstrating a direct interaction between them (Fig. 2c). To further investigate whether R9AP could directly mediate EBV infection, the localization of EBV and R9AP was examined in HNE1 cells overexpressing R9AP. Confocal microscopy revealed that Alexa Fluor 594-labelled-EBV co-localized with eGFP-tagged R9AP (R9AP-eGFP) on the cell membrane and in the cytoplasm (Fig. 2d). Furthermore, Biolayer interferometry (BLI) also con rmed a direct interaction between R9AP and gH/gL by demonstrating high-a nity binding of GST-R9AP 1-210 to His-gH/gL (K D =3.27E-09 M) (Fig. 2e). Taken together, these results demonstrated a strong, speci c, and direct interaction between R9AP and gH/gL. AMMO1 competes with R9AP for gH/gL binding Because AMMO1 inhibits EBV infection of both epithelial cells and B cells, we hypothesized that its binding site on gH/gL might overlap with that of R9AP. To test this directly, a competition binding assay was performed using BLI. As a control, we included the anti-gH/gL antibody CL59, which binds to epitope located in gH D-IV and blocks EBV infection e ciently in epithelial cells but only partially in B cells 25,37 . An appropriate amount of His-gH/gL protein was rst loaded on the Ni-NTA biosensor used for capturing His-tag-coupled protein. After a 30s of equilibration, the sensors were saturated with GST-R9AP or buffer intended for non-competitive control. The secondary binding of AMMO1 or CL59 to the biosensor was then monitored. To quantify the competition between each antibody and R9AP to the ligand gH/gL, raw signals of antibody binding were recorded with (Rc) or without (Ro) GST-R9AP association. Rc and Ro represent the responses of competitive binding and single ligand binding, respectively. The ratio of Rc/Ro re ects the intensity of competition. A ratio close to 1.0 indicates non-competitive binding and the competition could be observed for ratio <0.7. Generally, a smaller value indicates a stronger competitive binding between ligands toward the targeted protein. Our ndings showed that AMMO1 strongly competed with R9AP (Rc/Ro = 0.39) whereas CL59 did not (Rc/Ro = 1.02) (Fig. 2f).
To validate whether the interaction between gH/gL and R9AP could be blocked by gH/gL antibodies, coimmunoprecipitation was performed. gH/gL protein were pre-incubated with control IgG, AMMO1 or CL59 then mixed with R9AP. The results showed that AMMO1, but neither control IgG nor CL59, prevented the interaction between gH/gL and R9AP (Fig. 2g). Taken together, these results demonstrated that the AMMO1 overlaps with the R9AP binding site on gH/gL.

R9AP plays an essential role in EBV infection of both epithelial cells and B cells
Given above results, we further explored the role of R9AP in EBV infection in both epithelial cells and B cells. We rst knocked down R9AP expression by siRNAs in nasopharyngeal epithelial HNE1 cells. Two independent siRNAs targeting R9AP signi cantly reduced R9AP expression (Fig. 3a). Consistently, the number of EBV-infected cells, which expressed eGFP, showed a remarkable reduction among R9AP knockdown cells by the visualization of uorescence microscope (Fig. 3c). Meanwhile, the percentage of EBV-infected cells were determined by the ow cytometry analysis (Fig. 3a, b). The results showed that the e ciency of EBV infection was decreased at comparable levels in the R9AP knockdown cells. To exclude the potential siRNA off-target effects, we established R9AP genetic knockout HNE1 cells using CRISPR/Cas9. There was almost no detectable R9AP protein in the two sgRNAs (R9APsg1# and R9APsg2#) transfected HNE1 cells (Fig. 3d). Consequently, R9AP knockout almost abolished EBV infection in HNE1 cells (Fig. 3d). Importantly, the infection e ciency of R9AP knockout cells was mostly restored by ectopic expression of R9AP (Fig. 3e), indicating that the reduced EBV infection was not due to CRISPR off-target effect. R9AP knockdown in another nasopharyngeal epithelial HK1 cells also signi cantly inhibited EBV infection (Extended Data g. 3).
In addition to NPC, almost ten percent of gastric adenocarcinomas are associated with EBV infection. To examine the role of R9AP in EBV infection of gastric epithelial cells, we knocked down R9AP in the gastric epithelial MKN74 cells (Fig. 3f) and found that MKN74 cells with reduced R9AP expression were less sensitive to EBV infection (Fig. 3f).
As B cells are among the primary host cells for EBV, we further examined the function of R9AP in EBV infection to B cells. To overcome the low transfection e ciency of siRNA in B cells, we established R9AP genetic knockout Raji cells with lentivirus expressing the same sgRNAs used in HNE1 cells and found that both sgRNAs reduced R9AP expression (Fig. 3g). Consistently, R9AP knockout inhibited EBV infection of Raji cells by 70-80% compared to the empty vector control (Fig. 3g). Taken together, these data indicated that R9AP is essential for EBV infection in both epithelial cells and B cells.

R9AP overexpression promotes EBV infection
We then examined whether overexpression of R9AP could increase EBV infection. We found that EBV infection was signi cantly increased in nasopharyngeal epithelial CNE1 cells and gastric epithelial AGS cells overexpressing R9AP (Fig. 4a). Overexpression of R9AP also increased EBV infection of EBVnegative Akata cells (Fig. 4b). Taken together, these results indicated that R9AP overexpression enhanced EBV infection of both epithelial and B cells.

R9AP mediates EBV fusion
Since AMMO1 inhibits the fusion of viral and cellular membranes, we further examined whether R9AP was required for fusion. Using cell-based EBV fusion assay, we found that EBV fusion was decreased in R9AP knockdown cells (Extended Data g. 4a) but was increased in R9AP overexpressing cells (Extended Data g. 4b). These results suggested that R9AP also affected the fusion of viral and cellular membranes.
The transmembrane, Habc Tri-helical Bundle domain and linker of R9AP are required for EBV infection R9AP is a single-pass type transmembrane protein 38 . The wild-type (WT) R9AP includes 234 amino acids. According to the UniProt prediction 39 , the R9AP transmembrane domain includes amino acids 211 to 231. Thus, we constructed truncated mutants that deleted the transmembrane domain and the three amino acids in C-terminal end (R9AP 1-210 ) or deleted only the amino acids in C-terminal end of R9AP (R9AP 1-231 ) (Fig. 4c). We then expressed the WT and the two truncated mutants in HEK-293T cells (Fig.  4c). Deletion of the 3 amino acids in the C-terminal end of R9AP did not affect EBV infection compared to WT (Fig. 4d). However, deletion of the transmembrane domain plus the C-terminal three amino acids remarkably decreased EBV infection compared to WT, indicating that anchoring of R9AP in the cellular membrane is crucial for EBV infection (Fig. 4d). The N-terminus of R9AP includes Habc putative trihelical bundle domain, linker and SNARE homology domain 39,40 . We further determined which domains of R9AP were responsible for EBV infection by creating a series of R9AP deletion mutants (Fig. 4e). Deletion of amino acids 1 to 50 or 101 to 152 in the Habc putative trihelical bundle domain and linker impaired the function of R9AP in allowing EBV infection. On the other hand, deletion of either amino acids 51 to 100 in the Habc putative trihelical bundle domain and linker or SNARE homology 153 to 200 did not affect EBV infection (Fig. 4f). The remarkable decrease in EBV infection of cells expressing R9AP lacking amino acids 1 to 50 compared to the WT indicated that the Habc Tri-helical Bundle domain was essential for EBV infection.
R9AP was assumed to anchor RGS9 to the disk membrane 38 . We used online software InterPro and TMHMM to predict the sub-cellular localization of R9AP 41,42 . Both websites predicted the N-terminal of R9AP as extracellular localization (Extended Data g. 5a, b). To characterize the subcellular localization of R9AP, we expressed N-terminal Myc tagged amino acids 1 to 210 of R9AP (Myc-R9AP 1-210 ) or fulllength R9AP (Myc-R9AP FL ) in HNE1 cells then stained them using anti-Myc antibodies, with or without permeabilization using Triton X-100 (TriX100). Without TriX100 treatment, only cells expressing Myc-R9AP FL were stained, not those expressing Myc-R9AP 1-210 (Extended Data g. 5c). However, both cells expressing Myc-R9AP FL and those expressing Myc-R9AP 1-210 were stained after TriX100 treatment (Extended Data g. 5c). Additionally, we observed endogenous R9AP by immuno uorescence staining with anti-R9AP antibody which targeting the N-terminus of R9AP in non-permeabilized HK1 cells (Extended Data g. 5d). Together, these results indicated that N-terminal domain of R9AP could be exposed to cell surface.
Furthermore, we designed a biochemical experiment to con rm the extracellular localization of the R9AP's N-terminus by inserting a PreScission protease (PSP) -recognition site between the FLAG tag and Nterminal end of R9AP 1-210 (FLAG-psp-R9AP 1-210 ) or N-terminal end of R9AP FL (FLAG-psp-R9AP FL ). When PSP was added outside xed cells, the FLAG tag was removed only if the PSP recognition site was presented outside of cells, while not affected the binding site to the antibody against N-terminal of R9AP (Extended Data g. 5e). Western blotting analysis showed that HNE1 expressing FLAG-psp-R9AP 1-210 presented similar amounts of truncated R9AP in the presence or absence of PSP, detected by either anti-FLAG or anti-R9AP antibodies, indicating the intracellular localization of the PSP recognition site in this truncated mutant. However, PSP treatment dramatically reduced FLAG-psp-R9AP FL detection by anti-FLAG but not by anti-R9AP antibody (Extended Data g. 5f), demonstrating that the N-terminal FLAG tag of FLAG-psp-R9AP FL could be removed by PSP digestion and indicating that N-terminus of R9AP could be exposed outside the cells. Collectively, these results indicated that N-terminal domain of R9AP could locate at cell surface.

R9AP peptide impairs EBV infection in vitro and in vivo
Our results demonstrate that the N-terminal amino acids 1 to 50 of R9AP are essential for EBV infection.
Next, we evaluated the in vivo effect of R9AP [19][20][21][22][23][24][25][26][27][28][29][30] peptide against EBV infection in humanized mice. It was reported that SCID mice developed post-transplant lymphoproliferative disorders (PTLD) after reconstitution by intraperitoneal human peripheral blood mononuclear cell (PBMC) injection followed by EBV infection 43,44 . Previous studies also found that cord blood cells infected with EBV in vitro induced lymphomas when injected into NSG mice 45,46 . We reconstituted B-NDG mice by intraperitoneal injection of cord blood cells. Subsequently, EBV and peptide were simultaneously injected into the mice through their tail vain on day 0. Then, the mice were intraperitoneally administered with R9AP 19-30 peptide or control peptide at 20mg/kg on days 3, 7 and 14 (Fig. 5d). The engrafted T cells were able to slow the growth of EBV-induced lymphomas in a Cord Blood Humanized-Mouse Model 46 . To inhibit engrafted T cell function, T cell depleting monoclonal antibody (OKT3) was injected into the mice on day 7 after EBV injection. The peripheral blood of infected mice was collected at indicated time. EBV DNA in the peripheral blood was rst detected 4 weeks after inoculation and was observed to increase rapidly. EBV copy number was much lower in mice treated with R9AP 19-30 peptide compared to those treated with the control peptide on the 6th week after injection (Fig. 5e). The death of mice rst appeared on day 43 after EBV injection. Mice treated with R9AP 19-30 peptide survived longer than those treated with control peptide (Fig. 5f). However, there was no apparent difference in body weight between mice treated with the control and those treated with the R9AP 19-30 peptide (Extended Data g. 6a). Human CD20 and the latent EBV non-coding nuclear RNAs (EBERs) were detected in the spleens of all mice (Extended Data g. 6b), which indicated the successful construction of EBV infecting B-NDG mice model. However, the proportion of EBERs-positive cells in R9AP 19-30 peptide treated mice was much lower than the control peptide treated mice (Extended Data g. 6c). These data implied that R9AP 19-30 peptide protected mice against EBV infection in vivo and that targeting R9AP could be a potential strategy against EBV infection.

The expression of R9AP in tissues susceptible to EBV infection
Primary EBV infection is commonly thought to occur in vivo in the oropharynx, after which the virus spreads throughout the lymphoid tissues to infect B cells 47 . To determine the expression of R9AP in these tissues, we performed immunohistochemical (IHC) staining for R9AP in human tongue, oor of the mouth and lymphoid tissues. Human retina and liver tissues were used as positive and negative control, respectively 40,48 . As expected, R9AP mainly was found in the outer segment layer of retina, while not detected in the liver tissues (Extended Data g. 7). We observed that R9AP-positive cells mainly distributed in the basal layer of strati ed squamous epitheliums in the tongue and oor of the mouth as well as in follicular areas of lymphoid tissues in all ve investigated specimens of each tissue (Fig. 6a). Four of 5 pseudostrati ed ciliated columnar epithelium of the nasopharynx stained weakly. Dysplastic nasopharyngeal epithelium tissues were found to become susceptible to EBV infection compared to the normal nasopharyngeal epithelium tissues 47,49 . Consistently, 3 of 4 dysplastic nasopharyngeal epithelium tissues stained strongly for R9AP (Extended Data g. 8). However, R9AP was undetectable in all 5 of normal glandular epithelium of the stomach tissues (Extended Data g. 8). It was noteworthy that R9AP and EBERs could be detected, in similar distribution patterns in consecutive sections, in all analyzed EBV-associated human tumor samples, including 5 NPC specimens, 3 EBV-positive gastric carcinoma specimens, and 4 EBV-positive B cell lymphoma specimens (Fig. 6b).

Discussion
As a member of the herpesvirus family, EBV requires the coordination of a set of envelope glycoproteins and corresponding cellular receptors or co-receptors to e ciently infect host cells. To identify membrane-associated proteins that are important for EBV infection, we used unbiased transcriptomic microarray analysis combined with a siRNA library screening based on a highly e cient EBV infection model that we previously reported 27 . As far as we know, we identi ed R9AP as the rst characterized receptor to mediate EBV infection of both epithelial cells and B cells, which are the major cell types infected and potentially transformed to malignancy by EBV. Enveloped virus entering host cells can be divided into two steps: virus rst binding to the cell surface, then fusing with cell membrane and penetrating to release the capsid and nucleus into cytoplasm 11 . It was generally believed the B cell infection was initiated by EBV gp350 attachment to CD21, possibly allowing gp42 interaction with HLA class II. This close membrane protein interaction would trigger core fusion complex gH/gL and gB to ful ll nal viral-host membrane fusion 31 . We previously identi ed EphA2 and NRP1 as gH/gL or gB receptors for EBV fusion in epithelial cells. It remained unknown whether there are gH/gL or gB receptors for EBV fusion in B cells. Our results demonstrated that R9AP mediated EBV fusion through its interaction with gH/gL, which helps to fully understand the fusion mechanism. As mentioned above, several host factors have been found during different steps of EBV infection in either B cells or epithelial cells. Further studies will be needed to elucidate the relationship between R9AP and these host factors during EBV infection.
The competition between AMMO1 and R9AP to gH/gL binding facilitate the delineation of the mechanism of AMMO1-mediated neutralization. Previous study found that EBV lacking gH failed to infect both epithelial cells and B cells 50 , which indicates the indispensable role of gH in EBV infection. As expected, the human neutralizing antibody AMMO1, which binds to gH/gL D-I to D-II, can block EBV infection of both B cells and epithelial cells 34,35 , but the precise mechanism of neutralization was unknown. These results suggest a potential gH/gL receptor that might be used during EBV infection of epithelial cells and B cells. However, a gH/gL receptor in B cells or shared by epithelial cells and B cells remains unidenti ed for a long time. Our results found the interaction between R9AP and gH/gL was speci c and could be inhibited by AMMO1, which suggests that this is the mechanism of AMMO1mediated neutralization. These ndings also support the notion that vaccines designed to elicit antibodies targeting gH/gL and thus blocking the interaction between gH/gL and R9AP could raise the possibility to prevent EBV infection in both epithelial and B cells, thus reducing the incidence of EBV associated diseases. immuno uorescence staining and immunohistochemistry staining was purchased from Sigma-Aldrich (HPA049791, Germany). AMMO1 and CL59 antibodies were a kind gift from Professor Andrew T. McGuire (University of Washington) and Professor Richard Longnecker (Northwestern University), respectively.

R9AP is a potential target against EBV infection. Acute EBV infection causes infectious mononucleosis, EBV also reactivates in a subset of individuals with cancers, autoimmune diseases and
PreScission protease (PSP) was a kind gift from Professor Song Gao (Sun Yat-sen University Cancer Center). All other reagents were obtained from Sigma-Aldrich, unless indicated otherwise.
Microarray analysis. NPECs-Bmi1 MLCs and SLCs were formed, then, total RNA was extracted using TRIzol reagent (T9424; Sigma-Aldrich, Germany), according to the manufacturer's instructions. The integrity of the RNA was checked on a Bioanalyzer 2100 (Agilent Technologies, USA). The microarray experiments were performed by Shanghai Biochip Corporation (Shanghai; China) using the Agilent Whole Human Genome Oligo Microarray 4×44K (Agilent Technologies, California). The cRNA synthesis, cRNA labelling, sample fragmentation, and microarray hybridization were performed based on the manufacturer's standard protocols. The arrays were scanned using the Agilent Scanner G2505B (Agilent Technologies, California). Feature Extraction software (version 9.5.3.1, Agilent Technologies) was used to obtain raw data, which were normalized using the quantile algorithm in Genespring (version 9, Agilent Technologies). Differentially expressed genes were then selected according to the threshold set as fold change > = 2.0 and a P value < 0.05 according to the t-test. The identi ed differentially expressed genes were then subjected to GO (Gene Ontology) and KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway enrichment analysis. The microarray data were deposited with the NCBI GEO depository under the accession number of GSE159958. The zipped data includes raw data from the microarray for each sample, the normalized expression matrix of genes from all samples, and per-sample metadata. Differentially expressed genes were selected according to the threshold set as fold change >2.0 and a P value <0.05 by t-test. The bioinformatics analysis of differentially expression genes were conducted by GO (Gene Ontology) and KEGG (Kyoto Encyclopedia of Genes and Genomes) analysis. The heatmap was generated using GraphPad Prism software.
Gene silencing. ON-TARGET plus SMART pool siRNAs against 72 upregulated membrane-associated genes and ON-TARGET plus siCONTROL Non-Targeting pool siRNA which was used as control were all purchased from GE Dharmacon (California). The two single siRNA duplexes against R9AP were as follows: R9APsi1#:5'-GCGAGAUGAUCGACAACAU-3'; R9APsi2#:5'-GCAAAAGACGCGCCAGAAG-3'; All the siRNAs were delivered using RNAi MAX (13778150; Invitrogen, California) according to the kit instructions.
Expression of cDNA. The indicated plasmids were delivered using Lipofectamine3000 following the instructions. The establishment of R9AP stable expression EBV-negative Akata cells was based on the pBABE-Puro Retroviral system (RTV-001-puro, Cell BioLabs, California).
In vitro EBV infection. The utilized EBV containing eGFP (EBfaV-GFP), was prepared in EBV-positive Akata cells 36 . EBV production, labeling, and determination of MOI and infectious particles were performed as reported previously 27,28 . SLCs of NPECs-Bmi1 and HEK-293T were infected with EBV at an MOI of 300; HNE1, HK1, CNE1, AGS, MKN74, Raji and EBV-negative Akata cells cells were infected with EBV at an MOI of 1000. The indicated cell lines were incubated with EBV for 3h at 37°C, and unbound virus was discarded by washing with PBS three times. Then, cells were cultured in fresh medium for 24h, followed by quanti cation of eGFP positive cells using ow cytometry (cytoFLEX; Beckman).
Fusion assay. The virus-free cell fusion assay was performed according to previous published protocol (30). Effector HEK-293T cells were transfected with T7 luciferase reporter plasmids with a T7 promoter, gB, gH and gL. Renilla luciferase was used as internal control. To examine the effects on fusion after expressing R9AP, target HEK-293T cells were transfected with the T7 DNA polymerase expression plasmid together with the R9AP expression plasmid or empty vector. To test the effects on fusion after R9AP knockdown, target HEK-293T cells were transfected with R9AP siRNA or the control siRNA, followed by transfected with the T7 DNA polymerase expression plasmid 36h later. After 24h, 2.5× 10 5 effector HEK-293T cells were mixed with 2.5× 10 5 target HEK-293T cells in 24-well plates and cultured for another 24h. The luciferase activity was measured via a dual-luciferase reporter assay system (E2920, Promega, Wisconsin) using the GloMax-96 Microplate Luminometer. The ratio of re y luciferase activity to Renilla luciferase activity was used as the relative fusion activity.
Cells were treated with EBV at 37°C for 3h and subsequently washed three times by PBS. The percentage of EBV-infected cells was determined by ow cytometry at 24h post infection.
Protein expression and puri cation. GST-R9AP 1-210 was expressed in E.coli BL21 (AI) cells (CB105-02; Tiangen, China). The cells were cultured on TB plates containing 0.1mg/L of ampicillin (A010-10g; MDbio, China) at 37°C overnight. A single colony was picked and transfected into liquid TB medium supplemented with ampicillin for expansion until an appropriate OD value (0.6-0.8) was reached. The culture was then cooled down to 18 °C and isopropyl-thio-β-D-galactopyranoside (IPTG; 1122GR025; BioFroxx, China) was added to a nal concentration of 0.2mM to induce protein expression, which was continued for 8 hours at 18°C. The cells were then collected by centrifugation and resuspended in PBS, pH 7.4 containing Roche's complete Protease Inhibitor (EDTA-free) (11697498001; Roche, Switzerland). Then, the cells were lysed by pressured crushing at 4°C, after which the supernatant was collected by centrifugation and ltered through a 0.22μm pore-size syringe lter. The recombinant protein was puri ed using a gravity column with Glutathione Sepharose 4B resin (17075601; GE healthcare, Massachusetts) which was washed with the re-suspension buffer 3 times and eluted with PBS, pH 7.5 containing 10mM GSH (G4251-25G, Sigma-Aldrich, Germany) at 4°C. Then, the samples from cell re-suspension, lysate, supernatant, column ow-through, resin wash buffer and resin elution buffer were analyzed by SDS-PAGE to identify the collected protein. The elution fractions were combined and concentrated to 1mL using a 10kD molecular weight cutoff ultracentrifuge tube (UFC901096; Millipore, Germany) and loaded onto a size exclusion chromatography Superdex200 10/300GL column (28-9909-44; GE Healthcare, Massachusetts), which was eluted with PBS, pH 7.4 at room temperature. Fraction size was set to 0.5mL and each fraction was analyzed by SDS-PAGE. The SEC-puri ed protein was shock-frozen in liquid nitrogen and stored in -80 °C.
His-gH/gL was expressed in HEK-293T cells, which were transfected with the expression plasmid using polyethylenimine (PEI; 23966-1, Polysciences, Pennsylvania ) to HEK-293T cells at a ratio of 1mg plasmid/ 5mg PEI / 1×10 6 cells / 1L. The plasmid and PEI were diluted with HEK-293T medium (UP1000, Union, China) to 0.05mg mL -1 and 0.25mg mL -1 individually and mixed together for 15 minutes at room temperature before adding to adequate amount of cells. The transfected cells were cultured for 7 days under shaking at 120rpm. The supernatant was collected through centrifugation and ltered through a 0.22μm pore-size membrane. The following puri cation procedures were similar to those for protein expression in E.coli. His-resin (88229; Thermo, California) was used for a nity chromatography, washed by PBS pH7.4, 30mM imidazole (1460GR500; Biofroxx, Germany) and eluted by PBS pH7.4 (ZLI-9062; OriGene, China), 300mM imidazole. During SEC, PBS pH7.4 was used as running buffer for SEC.
Immunoprecipitation and GST pull-down assay. For the co-immunoprecipitation assay, HEK-293T cells were transfected with the indicated plasmid and lysed in lysis buffer containing 1%NP-40 (N885726; Macklin, China), 150mM NaCl (13423-6X1KG-R; Invitrogen, California), 2.5mM EDTA (E6758; Sigma-Aldrich, Germany), 20mM HEPES (H7523; Sigma-Aldrich, Germany) pH7.4 and protease inhibitor cocktail at 36h post transfection. The lysates were cleared by centrifugation at 12000 rpm and 4°C 10min . The supernatants were incubated with ANTI-FLAG M2 Gel (A2220, Sigma-Aldrich, Germany) or Anti-c-Myc Agarose A nity Gel (A7470, Sigma-Aldrich, Germany) overnight. Then, the gels with bound protein were washed three times with lysis buffer and subjected to WB analysis. For the GST pull-down assay, GST-R9AP 1-210 and His-gH/gL were incubated in lysis buffer overnight, then washed three times with lysis buffer and analyzed by WB. For the antibody competition binding assay, HEK-293T cells were transfected with plasmids encoding FLAG-R9AP or Myc-gH/gL for 36h, and lysed in lysis buffer. HEK-293T cells transfected with empty vector were used as control. Lysates containing Myc-gH/gL protein were incubated with 5mg IgG control, 5mg Ammo1, 10mg Ammo1, 5mg CL59 or 10mg CL59 overnight, and then incubated with lysates containing FLAG-R9AP or the control overnight. Finally, Myc-gH/gL was pulled down using Anti-c-Myc Agarose A nity Gel which was then washed three times with lysis buffer and subjected to WB analysis.
Biolayer Interferometry (BLI). BLI assays were performed on an Octet Red 96 instrument (18-1127;ForteBio, California) at 30°C with shaking at 1000rpm. All signals were recorded at standard frequency (5.0Hz). For kinetic analysis, Ni-NTA biosensors (18-5101; ForteBio, California) were incubated in PBS with 0.05% Tween20 (P7949; Sigma-Aldrich, Germany), the dilution buffer used throughout the whole assay, for 15min before performing the kinetic analysis. After 60s of primary baseline, His-gH/gL protein diluted with the buffer was loaded at 0.5mg mL -1 for 120s, followed by a secondary baseline equilibration for 30s. Then, the association of baseline-control and GST-R9AP 1-210 at a gradient of concentration from 6.25nM to 100nM was recorded for 180s, followed by a transition to a dissociation process for 600s and multiple rounds of regeneration with 10mM Glycine pH1.5 (GE healthcare). Similar procedures were performed for determination of binding a nity of R9AP peptide/control peptide to gH/gL, except changes of association time and dissociation time to 100s and 200s.The raw curves were baseline-subtracted before tting to the 1:1 binding model using the ForteBio data analysis software, after which the mean kinetic parameters (kD, kon, koff etc.) were rendered via a global t to all binding curves. For competition analysis, His-gH/gL protein diluted with the same buffer was loaded onto the Ni-NTA biosensor at 1μg mL -1 for 180s. After 30s of equilibration, primary association of R9AP or PBST was recorded until saturation for 600s, followed by the secondary association of AMMO1 or CL59 for another 600s. The concentrations of GST-R9AP 1-210 , AMMO1 antibody and CL59 antibody used in the assay were 200nM, 100nM, and 100nM, respectively. The sensors were regenerated with 10mM Glycine pH1.5. Real-time binding was recorded during the experiment and competitive/non-competitive behavior was determined by the binding response presented by different association couples.
Immuno uorescence staining. For the co-localization assay of R9AP with Alexa Flour 594-labelled EBV, HNE1 cells were transfected with a plasmid expressing eGFP-tagged R9AP for 24h, then co-incubated with EBV labeled with Alexa Fluor 594 (R37117, Molecular Probes, California) for 1h at 37°C. After removing the unbound virus, cells were xed with 4% paraformaldehyde (N1012, NCM biotech, China) in PBS for 20min at room temperature, and permeabilized with 0.1% Triton X-100 (0219485480; MPbio, California). The cell nuclei were counter-stained with 0.1% DAPI (D9542; Sigma-Aldrich, Germany). For the determination of exogenous R9AP localization, HNE1 cells were transfected with the indicated plasmids for 36h, then xed with 4% paraformaldehyde in PBS, and either permeabilized with 0.1% Triton X-100 or left untreated. Then, the cells were incubated with an antibody speci c for the Myc tag (M5546; Sigma-Aldrich, Germany) and an Alexa Flour 594-labelled goat anti-mouse IgG antibody. The cell nuclei were counter-stained with 0.1% DAPI. To detect the endogenous R9AP localization, HNE1 cells were xed with 4% paraformaldehyde in PBS and incubated with an antibody targeting R9AP (HPA049791, Sigma-Aldrich, diluted 1:100) and Alexa Flour 594-labelled goat anti-mouse IgG antibody. The cell nuclei were counter-stained with 0.1% DAPI.
PSP digestion assay. Cells in 12-well plates were transfected with the indicated plasmids for 24h, and then xed in 4% paraformaldehyde. Then, 10μg PSP was diluted in 500μl PBS and incubated with the xed cells at 4°C for 8 h. The treated cells were washed three times with PBS and collected for WB analysis.
In vivo EBV infection of humanized mice. Ten immunode cient B-NDG mice were purchased from BIOCYTOGEN (China), and divided into two groups. Human cord blood was obtained from Guangzhou Women and Children's Medical Center (China) after obtaining informed consent. Human cord blood mononuclear cells were separated using Ficoll-Hypaque density gradient, and then 1×10 7 cells were injected i.p. into 4 to 5-week-old B-NDG mice. At the same day, mice were injected through the tail vein with R9AP 19-30 peptide or control peptide at 20mg/kg of body weight, together with 30,000 infectious EBV particles. Then, the mice were injected intraperitoneally with R9AP 19-30 peptide or control peptide at 20mg/kg of body weight on days -3, -7 and -14, and injected intraperitoneally with 50 mg OKT3 (B104; Nobimpex, Germany) on day -7. Blood was collected from the mice to extract DNA from mice on days-0, -14, -28 and -42.
Measurement of EBV titers in mouse blood. To quantify the EBV DNA copy number in mouse blood, qPCR was used to detect the BamHI-W fragment of the EBV genome using the primers 5'-CCCAACACTCCACCACACC-3' and 5'-TCTTAGGAGCTGTCCGAGGG-3'. A calibration curve was made based on the EBV Nucleic Acid standards (BDS).
Analysis of R9AP and EBERs expression in human tissues. Samples of human tissue from the tongue, oor of the mouth, lymphoid tissuse, nasopharyngeal carcinoma, gastric carcinoma and B cell lymphoma were all obtained from patients who were admitted to the Sun Yat-sen University Cancer Center, and signed written informed consent forms. Samples of tissue from the tongue, oor of the mouth, and lymphoid tissues were analyzed by hematoxylin-eosin staining (H E) and R9AP antibody staining. To detect R9AP, the rabbit polyclonal antibody against human R9AP (HPA049791, Sigma-Aldrich, diluted 1:50) was used as primary antibodies, which was incubated overnight at 4°C. After washing three times in PBST, the tissue sections were incubated with anti-rabbit secondary antibody (1:1000, Zymed, California), and then treated with 3-diaminobenzidine tetrahydrochloride for 10 seconds, followed by staining with 10% Mayer's hematoxylin (ZSGB-Bio). Staining of para n-embedded tissue sections for EBV EBERs was performed using the ISH detection Kit (ISH-7001, ZSGB-Bio, China).
Analysis of the spleen of EBV-infected mice. After the death of all the ctrl-treated mice, the R9AP 19-30treated mice were euthanized. The spleens of all the mice were xed in formalin to examine if the animals had persistent EBV infection using H E staining, IHC staining with antibodies against human CD20 (B cell marker), and detection of EBV EBERs using the ISH detection Kit. The results were independently evaluated by two pathologists, who were blinded to the status of the samples. The expression of human CD20 and EBV EBERs was evaluated by counting 3 representative high-power elds (×40 objective)  vector, myc-gH, gL or myc-gB were cloned into the pCDNA6-Myc vector; for the GST pull-down assay and protein expression, the sequence encoding N-terminal amino acids 1 to 210 R9AP was cloned into the pGEX6p-1-GST vector, which was a kind gift from Professor Song Gao (Sun Yat-sen University Cancer Center), and the sequences of gL (23-137AA, M81 strain) and gH ectodomain (19-679AA, M81 strain) connected by a linker (GGGGS)x3 were cloned into the pCDNA3.1(+) vector; for the cell-based fusion assay, expression plasmids for pCAG-T7, pT7EMC-Luc, gB, gH or gL, which were a kind gift from Wolfgang Hammerschmidt (Helmholtz Zentrum Munich) and Professor R. Longnecker (Northwestern University), were used. Quantitative real-time PCR. Total RNA was extracted using TRIzol reagent (T9424; Sigma-Aldrich, Germany). To analyze gene expression, 1µg of RNA was reverse-transcribed using the RNA Reverse