Ephrin receptor A2, the epithelial receptor for Epstein-Barr virus entry, is not available for efficient infection in human gastric organoids

Epstein-Barr virus (EBV) is best known for infection of B cells, in which it usually establishes an asymptomatic lifelong infection, but is also associated with the development of multiple B cell lymphomas. EBV also infects epithelial cells and is associated with all cases of undifferentiated nasopharyngeal carcinoma (NPC). EBV is etiologically linked with at least 8% of gastric cancer (EBVaGC) that comprises a genetically and epigenetically distinct subset of GC. Although we have a very good understanding of B cell entry and lymphomagenesis, the sequence of events leading to EBVaGC remains poorly understood. Recently, ephrin receptor A2 (EPHA2) was proposed as the epithelial cell receptor on human cancer cell lines. Although we confirm some of these results, we demonstrate that EBV does not infect healthy adult stem cell-derived gastric organoids. In matched pairs of normal and cancer-derived organoids from the same patient, EBV only reproducibly infected the cancer organoids. While there was no clear pattern of differential expression between normal and cancer organoids for EPHA2 at the RNA and protein level, the subcellular location of the protein differed markedly. Confocal microscopy showed EPHA2 localization at the cell-cell junctions in primary cells, but not in cancer cell lines. Furthermore, histologic analysis of patient tissue revealed the absence of EBV in healthy epithelium and presence of EBV in epithelial cells from inflamed tissue. These data suggest that the EPHA2 receptor is not accessible to EBV on healthy gastric epithelial cells with intact cell-cell contacts, but either this or another, yet to be identified receptor may become accessible following cellular changes induced by inflammation or transformation, rendering changes in the cellular architecture an essential prerequisite to EBV infection.

Introduction Epstein-Barr virus (EBV) is a gammaherpesvirus that infects more than 90% of the world's population. While the virus is particularly well-known to infect B cells, causing infectious mononucleosis, Burkitt and Hodgkin lymphoma [1], it also infects epithelial cells and is associated with nasopharyngeal carcinoma and gastric cancer (GC) [2].
EBV-associated gastric cancer (EBVaGC) represents 8-10% of all GC cases worldwide, accounting for up to 80,000 cases per year [3,4]. EBVaGC was classified as one of four molecularly defined subtypes of GC, characterized by excessive cellular genome hypermethylation, frequent presence of PIK3CA mutations, overexpression of PD-L1/-L2 and CDKN2A silencing [3]. A causal role of EBV in this particular subtype of GC is expected due to the presence and clonality of the latent EBV episome in every cancer cell but not in surrounding tissue [5][6][7][8][9][10]. However, the exact sequence of events leading to EBVaGC is still a conundrum. Specifically, it is unclear whether viral entry into healthy cells initiates the pathogenic changes, or whether host cell modifications must precede the infection.
Virus entry into the main target cells is mediated by independent mechanisms; B cell infection requires the initial binding of the viral glycoproteins gp350 and gp42 to the B cell surface CD21 and HLA class II respectively (reviewed in [11]), followed by fusion mediated by viral envelope glycoproteins gH/gL and gB. In contrast, epithelial cell entry does not require gp350 or gp42. Instead, binding and fusion is directly mediated by gH/gL and gB and requires several entry receptors expressed on the epithelial cell surface, most prominently the ephrin receptor A2 (EPHA2) [11][12][13]. As members of the superfamily of transmembrane receptor tyrosine kinases, EPH receptors mediate short-distance cell-cell communication between neighboring cells upon binding of their ephrin ligands. EPH receptor-ephrin signaling plays a pivotal role during development but also in other cellular processes like adult stem cells niches, synaptic plasticity or bone homeostasis (reviewed in [14]). In primary epithelial cells, the location of EPH receptors is highly organized and restricted to cell-cell contacts [15,16]. It is unclear whether the receptor, tightly engaged in the junctions under homeostasis, is available for infection in healthy epithelium. Experimental studies of such questions have, however, been hampered due to the lack of a suitable primary cell model. Since the virus is strictly speciesspecific, there is no animal model [17] and most of the available studies have been performed on cancer cell lines. However, cancer cell lines markedly differ from in vivo tissue because they have accumulated mutations which alter cell organization, cell-cell contacts and signaling pathways. Here we use gastric adult stem cell-derived organoids [18,19] to analyze EBV infection of primary epithelial cells. In this culture system, human tissue-resident adult stem cells are seeded in an extracellular matrix and supplied with a mixture of growth factors. Stem cells proliferate and daughter cells differentiate to form 3-dimensional (3D) cell cultures, the organoids. To date, organoids can be grown from a vast variety of organs including the small intestine, liver, brain, prostate and stomach-each resembling the primary tissue they are derived from. They allow the study of a range of in vivo biological processes, including studies on viral and bacterial infection [20]. Of note, organoids have allowed the study of previously unculturable norovirus [21,22].
Our results support a role for EPHA2 in EBV entry into conventional cancer cell lines, as demonstrated previously. In contrast, organoids from normal, non-transformed tissue, are protected from viral infection, despite the expression of EPHA2. Subcellular localization analysis suggests that in normal organoids, EPHA2 is restricted to the cell-cell junctions which likely renders the protein inaccessible for the virus. Taken together, our results support the theory that cellular changes, probably in the cell architecture, are one prerequisite for infection.

Upregulated EPHA2 expression results in more efficient EBV infection in epithelial cell lines
Recent reports demonstrated the importance of EPHA2 expression for EBV infection in conventional cancer cell lines [12,13]. To enable a comparison of infection in organoids and cell lines, we first validated the infection in cell lines. Cell lines derived from the lymphocyte lineage-Akata, Raji and Elijah-hardly expressed EPHA2, whereas three epithelial cell lines tested-293, AdAH and AGS-highly expressed the receptor on RNA and protein level (Fig 1A  and 1B), confirming published data [12]. Using B cell-mediated transfer infection with Akata B cells containing a GFP-expressing EBV [23], the infection efficiency of the cell lines ranged from 6-26%, with highest infection rates in AdAH cells ( Fig 1C). EPHA2 expression increased approximately three-fold after addition of 10 ng/ml epidermal growth factor (EGF) for 24 h and could not be further enhanced with excess of EGF ( Fig 1D). The percentage of EBVinfected epithelial cells was doubled in EGF-pretreated AdAH cells compared to control cells (Fig 1E and 1F), corroborating previous results [13]. EPHA2 overexpression (Fig 1G and 1H) resulted in~two-fold higher EBV infection efficiency of AdAH cells (Fig 1I and 1J). Blocking of EPHA2 receptor either with EPHA2 ligand or an antibody reduced, but did not abolish the infection (Fig 1K and 1L).

EPHA2 is expressed heterogeneously in primary gastric epithelial cells
To evaluate, whether EPHA2 could also allow infection of EBV into primary cells, we examined expression of EPHA2 in organoids. Total RNA sequencing of six human gastric organoid lines [24] showed prominent expression of EPHA2 and other EPHA receptors (Fig 2A). RT-qPCR of several different organoid lines demonstrated that EPHA2 expression levels varied in a patient-dependent manner but overall resided in the same range as for epithelial cell lines ( Fig 2B). Organoid cultures grow under culture conditions that require 50 ng/ml EGF in the medium. Addition of higher EGF concentrations did not influence EPHA2 expression ( Fig 2C). Taken together, EPHA2 is expressed in primary gastric epithelial cells at a comparable level as in cell lines.

EPHA2 expression is not sufficient for EBV infection of human gastric organoids
Since the epithelial entry receptor EPHA2 is expressed in organoids, we assessed EBV infection efficiency in cells from human gastric organoids. For infection, cells from organoids were seeded in monolayers and infected via B cell-mediated transfer as described above for AdAH cells (Fig 3A). At 4-6 dpi fluorescence microscopy displayed a number of GFP-positive cells ( Fig 3B). However, control staining of CD45, a lymphocyte marker, showed that most GFPpositive (= EBV-positive) cells observed were remaining donor B cells (Fig 3C lower panel) and only single CD45-negative, GFP-positive epithelial cells could be detected (Fig 3C upper  panel). To quantify EBV infection efficiency, we performed flow cytometry analysis from infected organoid-derived monolayers ( Fig 3D). Staining with an anti-CD45 antibody confirmed that most GFP-positive cells were donor B cells (Quadrant 3) and less than 0.2% of cells could be detected in Quadrant 4, displaying newly infected epithelial cells (Fig 3E). Infection with cell-free virus yielded even lower numbers of infected cells (S1A Fig). To address the possibility that infection may be influenced by growth of the cells in 2D versus 3D, or infection via the apical versus the basolateral surface, we also performed microinjection of either cell-free virus, or donor B cells either to the apical or the basolateral side of the 3D organoids. Similar to that observed in 2D, neither parameter resulted in increased infection efficiency (S1B and S1C Fig). In summary, our data suggests that in organoids, expression of EPHA2 is not sufficient for EBV infection.

Efficient EBV infection of human gastric cancer organoids
Organoids can also be generated from gastric cancer tissue, enabling paired culturing of normal and cancer organoids from the same patient [19]. Notably, cancer organoids also display the cellular diversity of the patients' tumor [25] and biobanks of gastric cancer organoids contain all subtypes of cancers [26][27][28]. To explore the possibility to infect cancer organoids, we tested three EBV-negative cancer organoid lines ( Fig 4A).
Flow cytometry quantification demonstrated 1-9% of infected cells in the cancer organoids and an up to 70-fold increase of infection comparing normal organoids and cancer organoids ( Fig 4B). Immunofluorescence using control staining for CD45 to exclude remaining B-cells also confirmed infection in epithelial cells ( Fig 4C). Cancer organoids were also susceptible to infection with cell-free virus, although the efficiency was lower than with B-cell transfer, as expected (S2 Fig). Sorted GFP/EBV-positive cancer organoid cells could be selected using a previously introduced neomycin resistance cassette in the EBV bacterial artificial chromosome   [29]. After sort and 14 d expansion, cells were selected for 7 d in neomycin, picked and thereafter grown without selection pressure for over 6 months. In contrast to previous models using ex vivo cell lines or primary cells where the virus genome was lost over time [30][31][32], we obtained persistent infection (Fig 4D). EBER in situ hybridization initially confirmed the presence of the EBV small non-coding RNAs (EBERs) in this clonal line (Fig 4E). Conventional PCR was used to detect different EBV genes (Fig 4F). RT-qPCR confirmed EBV had established latent gene expression. This experimentally infected cancer organoid line exhibited mRNA expression of the virus maintenance protein EBNA1 together with expression of the latent membrane proteins LMP1 and 2a. However, the cells did not express the viral EBNA2, 3a, 3b or 3c mRNA, or indeed significant levels of mRNA of any viral lytic proteins, including BZLF1 ( Fig 4G). These data are consistent with the infected cells having a latency II phenotype. Taken together, in contrast to normal organoids, the cancer-derived organoids were able to maintain the viral genome and establish a long-term latent infection.

Distinct localization pattern of EPHA2 in normal organoids versus cell lines and cancer organoids
Blocking of EPHA2 either by ligand ephrinA1 or anti-EPHA2 antibodies reduced the infection in cancer organoids similar to that observed in cancer cell lines (Fig 5A, compare with Fig 1K), supporting a role for EPHA2 in EBV infection. Therefore, to analyze the molecular mechanism of the observed differences in EBV infection efficiency between non-transformed and transformed cells, we turned again to EPHA2. The differences in the EPHA2 mRNA expression and EPHA2 surface protein level were heterogeneous (Fig 5B and 5C). All normal organoid lines expressed about half of mRNA levels detected in the AdAH cell line. Cancer organoid lines expressed either similar, about 7-fold lower, or 2-to 3-fold higher mRNA levels as AdAH. We reasoned that if expression levels alone would influence infection, all cancer organoids would have higher expression. Because this was not the case, we searched for an alternative explanation.
Previous studies suggested that the localization of EPHA2 protein varies between epithelial cell lines and primary cells [15], pointing to the possibility that although expressed at similar levels, the protein may not be accessible for infection in primary cells compared to transformed cells. To test this hypothesis, we performed immunofluorescence (IF) analysis for EPHA2. In primary cells EPHA2 was located exclusively at sites of cell-cell contact ( Fig 5D) and co-localized with the adherens junction marker E-cadherin but not apical actin filaments (S3 Fig). Similarly, in healthy tissue and 3D organoids EPHA2 was localized predominantly at the cell-cell junctions (S4 Fig). In contrast, in cancer-derived organoids as well as cell lines, EPHA2 was not restricted to cell-cell contacts but also present at apical and basal sides of the cells, implying that this could possibly be an entry site (Fig 5E). These results support the hypothesis that if EPHA2 is an entry receptor, EBV might be unable to access it in non-transformed cells due to its sequestered localization (Fig 5F).

Number of EBV-positive cells in vivo is dependent on inflammation status of epithelium
To evaluate EBV infection in vivo in gastric tissue, we performed EBER in situ hybridization. Results illustrated that hardly any EBV-positive cells could be found in healthy gastric from the infected #30 cancer organoid line. The viral gene expression profile included expression of EBNA1, LMP1 and LMP2a plus the non-coding EBERs.
https://doi.org/10.1371/journal.ppat.1009210.g004  ) and (E) Immunofluorescence was performed for EPHA2. DNA was epithelium. However, EBV-positive epithelial cells were abundantly present in inflamed gastric tissue. As expected and corroborating previous reports, in EBVaGC, every cell stained positive for EBV (Fig 6A). Taken together, we conclude that healthy gastric epithelium is unlikely to be infected with EBV, despite the presence of EPHA2. Our data suggests that prior changes in the epithelium could be necessary to render EPHA2 receptor accessible or to promote other cellular changes to allow EBV to infect primary gastric epithelial cells (Fig 6B).

Discussion
EBVaGC accounts for over 8% of all GC cases [3]. The presence of a monoclonal virus genome in gastric epithelium from patients with GC as well as pre-malignant disease stages including chronic atrophic gastritis, but not healthy tissue [33], implies a causal role for EBV in the pathogenesis of gastric carcinoma. Yet, our understanding of how EBV initially accesses the gastric epithelium and establishes a persistent infection to drive the malignant changes in these cells remains poor.
In 2018, two independent papers determined EPHA2 as the EBV entry receptor in a variety of cancer cell lines including the gastric adenocarcinoma epithelial cell line AGS [12] and the nasopharyngeal carcinoma cell lines CNE1, CNE2 and HNE1 [13]. In these cells, EBV counterstained with Hoechst. Orthogonal view of the respective picture is depicted on the right. For separate channels, see S1 Fig  internalization and fusion was shown to be triggered by an interplay between EPHA2 and the viral entry glycoproteins gH/gL and gB. Interestingly, EPHA2 had likewise been demonstrated to function as an entry receptor into epithelial and endothelial cells for the Kaposi's sarcomaassociated herpesvirus (KSHV), the second human γ-herpesvirus besides EBV [34]. Again, entry is mediated by the highly conserved herpesvirus glycoproteins gH/gL and their interaction triggers EPHA2 phosphorylation and endocytosis of KSHV [35][36][37].
Using a similar set of epithelial cell lines, we confirmed that increased EPHA2 expression resulted in increased EBV infection, and correspondingly decreased EPHA2 accessibility due to ligand binding or blocking antibodies to EPHA2 resulted in decreased EBV infection. The range of effects that we observed (about 50% reduction of infection by blocking) matched the effects seen with these specific reagents in the published experiments. More efficient reagents may further increase the observed effects as published [12,13].
Also, although previously published experiments used cell-free virus, we used infection mediated by transfer from lytic B cells. It is possible that the dependence on EPHA2 may be reduced when virus is directly transferred in this manner, implying transfer infection may require additional receptors, such as integrins [38]. This alludes to the physiological relevance of transfer infection of epithelial cells. EPHA2 receptor is upregulated by inflammatory cytokines [39] and has also been shown to play a role in the recruitment of leukocytes to the site of inflammation [39][40][41]. Similarly, certain integrins are activated by inflammatory cytokines and also aid the recruitment of leukocytes. This suggests EBV-infected B cells (or lytic plasma cells) may themselves be recruited to sites of inflammation, aided by EPHA2 whilst making use of its expression to enter the epithelial cells. Indeed, such B cells are readily found in regions of intestinal inflammation such as inflammatory bowel disease and ulcerative colitis [42,43]. It has also been suggested that chronic inflammation could render epithelial cells sensitive to EBV infection (reviewed in [44][45][46]). Correspondingly, we identified EBV-infected epithelial cells in inflamed gastric tissue. Together, our data supports a function for EPHA2, but also points to the requirement for another receptor(s), such as integrins or another, yet unidentified receptor.
Our results in organoids indicate that in healthy gastric epithelial cells EPHA2 is located predominantly within the cell-cell junctions. This suggests that other events must precede the infection to allow accessibility of EPHA2 such as chronic inflammation or tissue damage which is known to alter the localization/expression of cellular receptors. However, our results do highlight that EBV infection of healthy gastric epithelial cells is likely to be an extremely rare event, which is unlikely to result in the establishment of a persistent latent infection.
Generally, EPHA2 is highly expressed in epithelial cells [47]. The role of the EPH system in cancer is complex. Up-as well as down-regulation of EPH expression had been reported in the literature (reviewed in [48,49]). Because the results from cancer cell lines indicated that expression levels of EPHA2 influenced infection-a two-fold to five-fold increase of expression resulted in double the numbers of infected cells-we compared the expression levels in cancer cell lines and organoids. Cancer organoids were infected with strikingly higher efficiency than the normal lines. In line with this, EBER in situ hybridization studies showed a lack of EBV infection of non-malignant epithelial cells [6,[48][49][50]. However, comparing cancer organoid lines from three patients, expression levels of EPHA2 did not fully mirror infectivity, and one cancer organoid line with very low expression level of EPHA2 was also infectable. This indicates, that additional factors other than mere expression of EPHA2 may be affecting EBV entry into epithelium.
One of the major differences between normal and transformed epithelium is the changes in cell-cell junctions triggered by epithelial to mesenchymal transition, whereby E-cadherin expression, amongst other junctional proteins, is downregulated resulting in a loss of polarity.
Indeed, downregulation of E-cadherin has been observed in tumors of epithelial origin (reviewed in [6,[50][51][52]), EBV-infected NPC [53], GC [54] and EBVaGC in particular [55]. Ecadherin and its associated complex dictates both the polarity and the motility of epithelial cells [56] and importantly, E-cadherin regulates the function and localization of EPHA2 [15,57]. Thus, a downregulation in E-cadherin expression might be expected to result in an altered localization of EPHA2. Indeed, EPHA2 exhibited a striking alteration in its localization on the cancer organoids and cell lines, exhibiting diffuse expression rather than being restricted to cell-cell junctions, as previously reported for transformed cells [15,57,58].
In contrast to the transformed cells, our normal organoids exhibited EPHA2 localization predominantly to cell-cell junctions [59,60] and EPHA2 co-localized with the prominent adherens junction marker E-cadherin [61,62], confirming earlier publications [15,16]. These results indicate that under physiologic conditions EBV infection of normal healthy epithelium may be limited by the inaccessibility of the entry receptor located within the cell-cell junctions. Other viruses such as adenovirus, coxsackie B virus or herpes simplex virus likewise use receptors in cell-cell junctions. The receptors were shown to be inaccessible to the virus unless junctions were disrupted (reviewed in [63]). Thus, we propose inflammation or pre-neoplastic changes as a prerequisite to render gastric epithelial cells susceptible to infection with EBV.
Taken together, we propose that the presence of EPHA2 on primary gastric epithelial cells in itself is not sufficient for efficient EBV infection. Other key drivers such as inflammation or somatic mutations likely changing the cell architecture are required to enable EBV infection, persistence and establishment of latency. The conceivable underlying mechanism might be a shift in the localization and hence accessibility of EPHA2 or another-yet to be identifiedreceptor. Subsequently, oncogenic hits induced by the virus could cooperate with previous events to accomplish malignant transformation towards EBVaGC.

Materials and methods
A detailed description of further materials and methods can be found in the online S1 Methods.

Ethics statement
Our study was reviewed by the ethical committee of the University Clinic, Wuerzburg, approval # 16/36. Human gastric tissue for the University of Birmingham study were obtained from the Human Biomaterials Resource Centre, Human Tissue Authority License number 12358, under the Research Tissue Bank ethical approval 20/NW/0001.
For 2D cultures, organoids were mechanically disrupted and seeded as pieces or single cells on conventional plastic (24-or 48-well plates) for infections or 8-well μ-slides for confocal microscopy (IBIDI, 80826). Single cell suspension was generated by treating mechanically disrupted organoids for 10 min with TrypLE Express (Gibco, 12605028) at 37˚C.

EBV B cell-mediated transfer infection
For B cell-mediated transfer infection, EBV+ Akata cells [29] were induced to activate the virus lytic cycle using 10 mg/ml human immunoglobulin (IgG) (MP Biomedicals, 0855049) for three days. In standard assays, 10 6 donor cells (induced Akata) were added to a~80%-confluent well of a 24-well plate that had been seeded 24 h before with 0.5-3.0 × 10 5 acceptor cells in the case of cell lines. Organoids seeded in 2D monolayers were usually cultured for 7-14 d until near-confluency before EBV transfer infection. After co-culturing for 24 h, donor cells were removed from acceptor cells by vigorous washing; transfer infection rate was assayed 72-96 h after the initiation of co-culture via fluorescence microscopy (Evos FL Imaging system) and flow cytometry (BD Accuri™ C6 Flow Cytometer) via the incorporated green fluorescence protein (GFP) gene in the EBV BAC. To differentiate between donor B cells and acceptor epithelial cells, additional staining with an APC-labelled CD45 antibody (Invitrogen, MHCD4505) was performed. Propidium iodide (PI) (Sigma Aldrich, P4864) staining was used to exclude dead cells.

Immunofluorescence (IF)
For IF of 3D organoids, organoids in Matrigel were resuspended in 500 μl cell recovery solution (Thermo Fisher Scientific, 12648-010) with a widened pipette, transferred to falcon tube and incubated for 45 min on ice until Matrigel was dissolved. For IF in 2D, dissociated gastric organoids were seeded onto 8 well μ-slides (IBIDI, 80826) to form 2D monolayers for immunofluorescence staining. Cells were grown for 7-14 d to reach approximately 90% confluency. Fixation was performed with 4% PFA for 20 min at room temperature, washed three times with PBS and permeabilized in 1 X PBS supplemented with 0.3% Triton-X, 1% DMSO and 1% fresh BSA for 1 h. Stainings were performed with primary antibodies (Occludin: Santa Cruz Biotechnology, sc-133256; CD45: Santa Cruz Biotechnology, sc-1178; EPHA2: Cell Signaling, 6997S; Pan-cytokeratin: Santa Cruz Biotechnology, sc-8018; E-cadherin: BD Bioscience, 610182) in 1 X PBS supplemented with 5% goat serum (Thermo Fisher Scientific, 31872) overnight at 4˚C followed by Alexa Fluor (AF) 488 or AF647-conjugated secondary antibodies (Cell Signaling) in 1 X PBS supplemented with 5% goat serum for 3 h at room temperature. Actin filaments were stained with Phalloidin (Thermo Fisher Scientific, A22283) and DNA was stained with Hoechst 33342. After washing three times with PBS, stained cells were visualized using a confocal microscope (Leica, TC5 SP5 X).

RT-qPCR for EBV gene expression
Absolute quantification of EBV gene transcripts by RT-qPCR was performed as previously described [68] and detailed in S1 Methods. All primer sequences are listed in S3 Table. Supporting information S1 Methods. A detailed description of further materials and methods.   Table. Organoid medium composition for human gastric organoids. ROCK inhibitor was added only after the initial seeding and passaging of the organoids. For basal medium (AD++), Advanced Dulbecco's modified Eagle medium (DMEM)/F12 supplemented with 10 mmol/l HEPES and GlutaMAX 1 X was used. CM: conditioned medium; inh.: inhibitor; N-Ac: N-acetylcysteine; EGF: epidermal growth factor; FGF-10: fibroblast growth factor-10; TGF-β: transforming growth factor-β; ROCK: Rho-associated coiled-coil forming protein serine/threonine kinase. (PDF) S3