Epstein-Barr Virus Infection in Ex Vivo Tonsil Epithelial Cell Cultures of Asymptomatic Carriers

Epstein-Barr virus (EBV) is found frequently in certain epithelialpathologies, such as nasopharyngeal carcinoma and oral hairyleukoplakia, indicating that the virus can infect epithelialcells in vivo. Recent studies of cell lines imply that epithelialcells may also play a role in persistent EBV infection in vivo.In this report, we show the establishment and characterizationof an ex vivo culture model of tonsil epithelial cells, a likelysite for EBV infection in vivo. Primary epithelial-cell cultures,generated from tonsil explants, contained a heterogeneous mixtureof cells with an ongoing process of differentiation. Keratinexpression profiles were consistent with the presence of cellsfrom both surface and crypt epithelia. A small subset of cellscould be latently infected by coculture with EBV-releasing celllines, but not with cell-free virus. We also detected viral-DNA,-mRNA, and -protein expression in cultures from EBV-positivetonsil donors prior to in vitro infection. We conclude thatthese cells were either already infected at the time of explantationor soon after through cell-to-cell contact with B cells replicatingEBV in the explant. Taken together, these findings suggest thatthe tonsil epithelium of asymptomatic virus carriers is ableto sustain EBV infection in vivo. This provides an explanationfor the presence of EBV in naso- and oropharyngeal pathologiesand is consistent with epithelial cells playing a role in theegress of EBV during persistent infection.

INTRODUCTION

Top

Introduction

Epstein-Barr virus (EBV) is a ubiquitous, persistent human herpesviruswhose tropism for B lymphocytes is well established. However,some reports have also suggested an epithelial-cell tropismfor the virus based on claims of having identified EBV-infectedpharyngeal epithelial cells in vivo (46). This led many investigatorsto propose that the mucosal epithelium of the nasopharyngealregion was the site of EBV persistence and that the infectionof B cells was not critical for viral persistence (1, 41, 42,61). Subsequently, however, these findings were challenged basedon the failure of other investigators to reproduce them (25,35) and the mounting body of evidence implicating memory B cellsas the site of EBV persistence (4, 13, 17, 51, 52). Specifically,it has not been possible to reproducibly detect infected epithelialcells in cross sections of tonsils or exfoliated epithelialcells in the saliva of acutely infected patients (25, 35). Eventhough these studies did not rule out low levels of infectedcells, they encouraged the belief that EBV does not normallyinfect healthy epithelial cells in vivo. Nevertheless, EBV is100% associated with naso- and oropharyngeal epithelial pathologies,such as nasopharyngeal carcinoma (NPC) (3) and oral hairy leukoplakia(OHL) (18). In NPC, the epithelial cells are predominantly latentlyinfected, whereas in OHL, an AIDS-associated lesion of the oralepithelium, infectious virus is produced by epithelial cellsin large amounts (58). From these findings, it is clear thatEBV can latently infect and replicate in oral- and nasopharyngealepithelia in vivo. This raises the possibility that these lesionsmay be amplified versions of infected healthy epithelium.

A recent in vitro breakthrough in understanding the infectionof oropharyngeal epithelial cells was provided when ß1integrin was identified as a crucial component of the receptorthat mediates EBV infection in subsets of oral epithelial celllines (54). Other in vitro findings provided a mechanistic explanationfor this proposed dual tissue tropism (8, 10). Virions producedby B cells in vitro have glycoprotein expression patterns thatfavor infection of epithelial cells and vice versa. This couldresult in a preferential exchange of EBV between the tissues.If EBV in vivo is transmitted back and forth between epithelialcells and B cells, the tissues must be in close proximity toone another. Tonsil reticulated crypt epithelium (37, 38) isone such site where EBV-infected B cells are in intimate contactwith neighboring epithelium (2, 34). Great improvements havebeen made in understanding the role of B cells in EBV persistencein vivo due to the accessibility of infected primary B lymphocytesin the blood and lymph nodes (51, 52) and the availability ofan in vitro infection system (26). This work has establishedthe importance of B cells in the tonsil lymphoepithelium asa critical site for the establishment of latent persistent infectionand for replication of the virus (51). In contrast, evidencethat nasopharyngeal epithelial cells play a role in EBV persistenceand/or replication in healthy carriers is still lacking.

One possible role for the epithelium in the nasopharynx is asthe first tissue that becomes infected after exposure to thevirus, because EBV is transmitted via saliva (19). Lytic replicationin epithelial cells could then act as a transit route or anamplification step prior to infection of the main target, Bcells residing in the underlying lymphoepithelium. Alternatively,epithelial cells could be used to transit and release virus,produced from infected B cells in the lymphoepithelium, backinto saliva. This kind of dual tissue tropism strategy is consistentwith the behavior of other ubiquitous herpesviruses (43), andin vivo evidence for such an infection route has recently beenprovided for simian immunodeficiency virus in monkeys (50).

The failure to detect infected epithelial cells in vivo andthe lack of an accurate primary in vitro nasopharyngeal epithelialculture model has greatly hampered research in understandingthe role these cells might play in persistence. For these reasons,we set out to establish and characterize a primary tonsil epithelialculture system that retains the heterogeneity of the tissuefound in vivo. We have developed early ex vivo tonsil epithelial-cellcultures that express keratin markers indicative of crypt andsurface tonsil epithelia. A subset of these cells is reproduciblyinfectible in vitro by coculture with EBV producer cells. Surprisingly,evidence of EBV infection was detected prior to the additionof exogenous virus in cultures from EBV-positive tonsils evenwhen maintained in the presence of acyclovir. These observationssuggest that the epithelial cells were infected either in vivoor rapidly after explantation by virus already present in thetissue explants. To our knowledge, these findings are the firstto describe EBV infection of healthy tonsil epithelial cellsderived from asymptomatic virus carriers. Our data support themodel of dual epithelial-lymphoid tropism for the virus in vivoand the possibility that healthy tonsil epithelium in vivo mayplay a role in transmission of the virus as part of the virallife cycle and suggest that EBV can play an initiating rolein associated epithelial lesions, such as NPC and OHL.

MATERIALS AND METHODS

Top

Materials and Methods

Primary cultures and cell lines. Tonsillectomies were performed at Massachusetts General Hospitalon 6- to 14-year-old patients because of obstructed-breathingdisorders. Tonsils were kept on ice in PBSAPSF (1x phosphate-bufferedsaline [PBS]-0.5% bovine serum albumin-PennStrepFungizone; Gibco)and were processed between 2 and 6 h after surgery. The epithelialculture method we applied is a modified version of the methodfor culturing epidermal keratinocytes developed by Lindbergand Rheinwald, Rheinwald and Green, and Schon and Rheinwald(30, 40, 44). Epithelial sections of the tonsil exterior wereremoved with surgical tools. Connective tissue sections of theinterior were separated from soft tissue (lymphocytes) by scrapingthe blunt side of a razor blade over the tissue sections. Thetissue sections were washed in cold PBSAPSF at least 10 times,minced with razorblades, and extensively washed in cold PBSAPSFto remove nonepithelial stroma (mostly lymphocytes) and debris.The lymphocytes were isolated from the supernatant by Ficoll-Hypaque(Amersham Pharmacia) centrifugation, washed two times with 1xPBS-0.5% bovine serum albumin, and resuspended at 2 x 10⁷/ml.Small explants of 1 to 2 mm³ were resuspended in 10% fetal bovineserum (FBS)-PennStrepFungizone-Dulbecco's modified Eagle's medium(DMEM)-F12-1 mM calcium (Gibco) (high-calcium serum-containingmedium), washed twice, and placed in a 100-mm-diameter dishand incubated overnight at 37°C in an incubator (day zero).This dramatically reduced the occurrence of fungal and/or bacterialcontamination. For ex vivo infection experiments, 0.1 mM acyclovir(UDL Labs, Rockford, Ill.) was added to the medium. The nextday, tissue sections were washed in medium again to remove theremaining lymphocytes, and transwell devices (24-mm diameter;Costar) were placed on top of $~$ 30 to 40 explants. An inert sterilestirring magnet was placed in the transwell for two reasons:to keep the explants attached to the bottom of the plate andto stimulate epithelial-cell migration out of the explants.Serum-containing medium was added in and surrounding the transwells.Three transwells were placed in one 100-mm-diameter culturedish. The medium was replaced on day 2. On days 3 and 4, dependingon individual culture growth, the transwells were gently removedand discarded. The medium was replaced by defined keratinocyteserum-free medium (SFM) plus keratinoctyte growth factor (KGF)(Invitrogen) containing 0.1 mM Ca. Most tissue explants remainedconnected to the bottom of the plate. Fresh medium was addedeach day after gentle washing. Typically, around days 7 to 10in culture, colonies 1 cm in diameter surround the tissue explants.If the explants were removed when the colonies were less than $~$ 0.2 cm in diameter, further colony growth was markedly curtailed.Tissue explants were gently removed by suction with a vacuumPasteur pipette. The serum-free culture medium was then changedevery 2 days. Culture dishes with contaminating spindle-likefibroblasts were discarded.

All cell lines were maintained under standard conditions with5% CO₂ at 37°C in a humidified atmosphere. IB4, B95-8 (agift of E. Kieff), BJAB, Akata, and Daudi (all American TypeCulture Collection) cell lines were maintained in complete RPMI1640 medium supplemented with 10% fetal bovine serum, penicillinplus streptomycin, and L-glutamine. EBfaV-GFP (green fluorescentprotein) cells were maintained in complete RPMI with 100 µgof G418 (Gibco)/ml. EBfaV-GFP (a gift of Richard Longnecker)is a derivative of the B95-8 virus-producing cell line in whichthe viral LMP2 gene is replaced by a gene cassette consistingof the enhanced-GFP gene driven by the cytomegalovirus promoterand the neomycin resistance gene driven by the simian virus40 promoter (48).

PCR. Reverse transcription (RT)-PCR for latent genes was performedas previously described (5, 20). Briefly, RNA was purified from2 x 10⁵ epithelial cells and 5 x 10⁶ lymphocytes using Trizolreagent (Invitrogen) as described by the manufacturer. Lymphocytesfrom EBV-negative tonsils, the EBV-negative BJAB B-cell line,and the EBV-positive IB4 B-cell line were used as controls.cDNA was prepared as described previously (5, 20), except thatthe 20-µl cDNA mixture was not ethanol precipitated butbrought up to 100 µl with high-performance liquid chromatographyH₂O and used directly. PCR was performed on the synthesizedcDNA for EBNA1U-K, EBNA1Q-K (EBNA1 from the Qp promoter), EBNA2,LMP1 (latent form), and LMP2a. The reaction was carried outin a final volume of 50 µl in a mixture consisting of50 mM KCl, 20 mM Tris (pH 8.4), 2.5 mM MgCl₂, 0.2 mM deoxynucleosidetriphosphates, and 20 pM (each) amplimers (final concentrations).The exception was LMP1, for which 3.0 mM MgCl₂ was used. Theamplimers and PCR strategies are described in detail elsewhere(5, 20). The PCR products were visualized by Southern blottingas described previously (33). Probes for blotting were madefrom PCR products derived from the IB4 cells.

Coculture and in vitro infection. EBfaV-GFP cells and B95-8 cells were treated with 20 ng of tetradecanoylphorbol acetate (Sigma)/ml and 3 mM butyric acid (Sigma) for1 to 4 days, while Akata cells were cross-linked with 100 µgof Fab immunoglobulin G (IgG) (Jackson Laboratories)/ml overnightto induce virus release. Cell suspensions were centrifuged (5min at 3,000 x g), and the supernatants were filtered througha 0.45-µm-pore-size filter. One milliliter of supernatant(cell-free virus) was added to 2 x 10⁵ epithelial cells in six-wellplates in 3 ml of medium. Daudi cells were incubated for 1 hin undiluted and 10-fold-diluted supernatants, washed, and placedin culture. For infection through cell-to-cell culture, producercells were stimulated overnight, washed, and resuspended inSFM. Virus-releasing cells were added to six-well plates andincubated with 1 x 10⁵ to 2 x 10⁵ day 7 to 10 subconfluent layersof primary epithelial cells at a 1:1 or 2:1 ratio. Floatingcells were removed at various time points (days 1 to 3) postinfectionby washes, and epithelial cells were transferred to cultureslides 4 days postinfection and prepared for staining.

Staining. Cells were cultured as described and replated at various timepoints (typically days 7 to 10) in two- and eight-well chamberedculture slides (Costar) in SFM. Cytospins of IB4 (an EBV-positivecell line) and BJAB (an EBV-negative cell line) cells were usedas controls for viral-protein detection and B-cell (CD21) andepithelial marker (Ber-ep4) expression. Cytospins and cultureslides were treated the same way in all staining procedures.For immunohistochemistry (HC), cells were fixed and permeabilizedfor 10 to 15 min with 100% chilled acetone. For immunofluorescence(IF), cells were fixed and permeabilized for 30 min at roomtemperature in 4% paraformaldehyde, followed by 10 min with0.1% Triton X-100 in phosphate-buffered saline (PBS) at 4°C.The slides were subsequently blocked with appropriate serumfor 1 h at room temperature prior to indirect staining. Forindirect IF staining of viral proteins, the cells were incubatedwith primary monoclonal antibodies: LMP1-S12 (dilution, 1:5,000),EBNA1, viral capsid antigen (VCA), and EAD (dilution, 1:1,000)(gifts of J. Middeldorp) and BZLF1 (dilution, 1:10) (Dako) atroom temperature for 30 min and then washed three times in PBS.Secondary goat anti-mouse Alexa 488 and Alexa 594 antibodies(dilution 1:5,000) (Molecular Probes, Breda, The Netherlands)were incubated for 1 h in the dark. The slides were washed withPBS (three times) and examined with a fluorescence microscope(Nikon E400). For HC and IF of cellular proteins, primary antibodiesto keratin 8 (K8), K18, K8/K18, K10, K19, K13, involucrin, fillagrin,CD21, and AE1/AE3 (pankeratins) (Neomarkers, Fremont, Calif.)were typically diluted 1:200 and incubated for 45 min at roomtemperature. For HC, the slides were developed with an LSABdetection kit (Dako) used according to the manufacturer's instructions,and the slides were visualized using a light microscope (NikonTE200). For IF, the slides were developed as described above.

RESULTS

Top

Results

Culture and characterization of primary tonsil epithelial cells. The culture process was performed in three phases. The firstculture phase, DMEM-F12 plus 10% FBS with a high (1 mM) calciumconcentration, is designed to stimulate the migration and adherenceof epithelial cells out of the small tonsil tissues to the culturedish. The second phase, SFM with low (0.1 mM) calcium plus definedKGF, ensures the elimination of lymphocytes and fibroblaststhrough cell death but promotes epithelial-cell viability andgrowth. In this phase, keratinocyte growth and migration arestill supported by the existing tissue explants (40). In thethird phase (SFM plus KGF), the explants are gently removedand the epithelial cells are replated to another dish at $~$ 60%confluency. Cells with clear epithelial morphology started tomigrate on days 3 to 5 in culture, and on days 5 to 8, coloniesof various sizes surrounded most individual tissue explants.The details are described in Materials and Methods. Figure 1shows a schematic and representative photogaphs of the threeculture phases. In general, the primary epithelial cells remainedviable for up to 4 weeks in culture.

View larger version (49K):
[in this window]
[in a new window]
FIG. 1. Tonsillar explant culture scheme. Cultures were grown as described in Materials and Methods. The procedure involves three separate steps. In step 1, tonsillar explants are placed beneath tissue culture transwell inserts. The membrane is porous, and a weight is placed on top of the insert membrane. DMEM-F12 containing 10% FBS and 1 mM Ca²⁺ was used during the first 4 days of culture. In step 2, the inserts are removed and the culture medium is changed to a low-calcium defined serum-free medium with keratinocyte growth factor (Gibco). This ensures the outgrowth of epithelial cells, abolishes fibroblast growth, and kills lymphocytes. In step 3, the explants are removed and epithelial cells are transferred to new dishes and culture slides. Photographs of the cultures at different stages are shown. Epithelial cells begin to migrate out of the tissue explants (T) on day 4 (d4) (black arrow). On day 7, large colonies surround the tissue explants with no visible indication of surviving lymphocytes. On day 7, large colonies (>1 cm in diameter) of cells with typical epithelial morphology surround the explants. By day 10, adherent cells have different morphologies, some cells having acquired a very large cytoplasm-to-nucleus ratio (white arrow), often seen in differentiated squamous epithelial cells. Thus, it appears that an active epithelial differentiation process is ongoing in the cultures.

To characterize the purity, heterogeneity, and differentiationstage of the cultured ex vivo tonsil epithelial cells, we performedHC and IF staining for epithelial and lymphoid markers. Day7 to 8 cultures were isolated and plated at $~$ 80% confluency onculture slides and used for staining the next day. One hundredpercent of the cells stained positive for AE1/AE3 (pankeratinocytemarker) but were completely negative for CD21 (pan-B-cell marker)(Fig. 2A, B, and C), indicating that the cultures were of $~$ 100%epithelial lineage. Similar results were obtained when the cellswere analyzed by flow cytometry. The absence of keratin 10 andfillagrin expression as early as day 7 and later (not shown)indicated that the cells had not undergone growth arrest andwere of the nonkeratinizing squamous epithelial lineage (32).Individual epithelial cell types can be distinguished basedupon expression of differentiation markers and morphology (30,32). A subset of morphological large epithelial cells appearedin the cultures over time (Fig. 1, day 7 versus day 10). Thesecells were epithelial surface antigen (ESA) negative (earlydifferentiation marker [6]), involucrin positive (late differentiationmarker [56]), and keratin 13 negative (early differentiationmarker) and had a large cytoplasm/nucleus ratio (Fig. 2E, I,and J). This expression profile is consistent with differentiatedsquamous epithelial cells (32, 56), suggesting that a processof terminal differentiation was ongoing in the cultures. Occasionallyepithelial cells tended to round up spontaneously and detachfrom the dish, a common behavior of terminally differentiatingprimary epithelial cells in vitro (9). Positive staining ofthese cells for involucrin confirmed that they were differentiated(data not shown).

View larger version (55K):
[in this window]
[in a new window]
FIG. 2. HC and IF analyses of primary tonsil epithelial cells indicate cellular heterogeneity. Primary tonsillar epithelial cells (days 7 to 10) were grown to confluence on chambered culture slides. Immunostaining was performed with a panel of monoclonal antibodies against epithelial (keratin, involucrin, and ESA) markers and a B-cell marker, CD21. Positive staining is red. For HC, counterstaining was performed with hematoxylin and eosin (blue.) For IF, nuclei are stained with DAPI (4',6'-diamidino-2-phenylindole) (blue). Larger cells (arrows) are negative for ESA (Ber-ep4 clone) and positive for involucrin. This confirms that the phenotypic differences are linked to the differentiation stage.

Most epithelial tissues are complex, and individual keratinprotein expression can be used to define the origin of epithelialcells (30). Tonsil epithelium is comprised of stratified surfaceand reticulated crypt epithelia (11, 38). In situ staining oftonsil epithelium suggests that expression of the simple epithelialkeratins K18 and K8 is limited to the crypt epithelial cells(11). We detected variable K8 and K18 expression early in ourex vivo tonsil cultures, suggesting that a subpopulation ofepithelial cells in our cultures were of crypt epithelial lineage(Fig. 2D and H). In addition, we detected variable K19 expression,another potential tonsil crypt lineage marker (11). The variableexpression of K19 in the cultures was particularly apparentby immunofluorescence staining compared with pancytokeratinexpression (Fig. 2A, B, F, and G) (11, 32). In summary, ourculture method produces heterogeneous squamous epithelial culturesin which differentiation appears to be ongoing and keratin expressionpatterns are consistent with those found in tonsil epithelialcells in situ (11, 32). These results are summarized in Table1.

View this table:
[in this window]
[in a new window]
TABLE 1. Summary of phenotypic characterization of primary epithelial cultures from the palatine tonsil

In vitro infection. If tonsil epithelium is important in EBV pathogenesis, the cellsshould be infectible with EBV. Coculture with EBfaV-GFP-expressingEBV-producing cells (49), but not incubation with infectiousviral supernatant, resulted in infection of all (10 of 10) independentepithelial cultures tested (Fig. 3). This is consistent withearlier findings that cell-free virus is usually incapable ofor inefficient at infecting most epithelial cells in vitro (16,23, 45). The infection frequency with EBfaV-GFP was estimatedto be $~$ 0.01 to 0.5% of the total cell population by countingGFP-expressing cells with epithelial morphology on day 3 postinfection.To test the possibility that the titer of cell-free virus wastoo low, we attempted to infect Daudi cells with the cell-freeEBfaV-GFP virus. We observed infection levels of up to $~$ 70% inDaudi cells with the same cell-free virus supernatant that wasunsuccessful at infecting primary epithelial cells (Fig. 3C).This indicates that susceptibility to infection, rather thanviral titers, is the limiting factor.

View larger version (68K):
[in this window]
[in a new window]
FIG. 3. Primary tonsillar epithelial cells are susceptible to in vitro infection with EBfaV-GFP by cell-cell contact but not with cell-free virus. (A) GFP expression could be detected in epithelial cells as early as 24 h postinfection (white arrows). Occasional adherent producer B cells are indicated with white arrowheads. Left panels, bright field; right panels, fluorescence for the same field of cells. (B) GFP-expressing cells are cytokeratin positive (indicated by white arrows). Left panels, red fluorescence for cytokeratins; right panels, green fluorescence for GFP expression on the same field of cells. (C) Infection by cell-free virus was never observed with epithelial cells, but Daudi cells could be efficiently infected (36 h postinfection) with EBfaV-GFP viral supernatant; the infection efficiency was dose dependent.

To prove that the infected cells were of epithelial origin,they were stained for cytokeratin expression (Fig. 3B and Fig.4A and B). True infected epithelial cells (Fig. 3A and B) couldbe distinguished from occasional adherent EBV producer lymphoidcells (Fig. 3A and B) based on their extremely bright GFP levels,morphology, and lack of keratin expression. Dual imaging forcytokeratins and involucrin suggests that both undifferentiatedand large differentiated epithelial cells could be infected(Fig. 4A to C).

View larger version (72K):
[in this window]
[in a new window]
FIG. 4. In vitro infection of primary epithelial cells. Seven- to 10-day epithelial cells were incubated in a 1:1 ratio for 1 to 3 days with viral producer B cells stimulated to produce virus. (A and B) Merged immunofluorescence images of EBfaV-GFP-infected epithelial cells are cytokeratin positive (red) and express GFP (green), which appears as yellow; nuclei are stained with DAPI (4',6'-diamidino-2-phenylindole) (blue). (C) Involucrin-positive cells are infected with EBfaV-GFP. (D to I) Primary tonsil epithelial cells infected in vitro with B95.8 EBV are latently infected. Isolated cells with epithelial morphology coexpress LMP1 (red) (D and G) and EBNA1 (green) (E and H), indicating a latent stage of infection in these cells. Nuclei are stained with DAPI (blue). Merged images (F and I) show the nuclear location of EBNA1 (light blue) and the cytoplasmic location of LMP1.

We next tested whether latent viral proteins are expressed andif susceptibility to infection is perhaps limited to the artificialEBfaV-GFP virus strain, which lacks LMP2 (49). We coculturedthe epithelial cells with phorbol myristate acetate-butyricacid-stimulated B-95.8 marmoset and IgG-stimulated Akata (notshown) cell lines and performed IF staining for cytokeratins,EBNA1, EBNA2, LMP1, and LMP2. Both virus strains were able toinfect epithelial cells, as judged by the detection of EBNA1and LMP1 protein expression within the same cells (Fig. 4D toI). EBNA2 and LMP2 expression was not detected. These resultsindicate that latent genes are expressed in the EBV-infectedprimary tonsil epithelial cells and that infection is not strainspecific.

Ex vivo infection. During the course of characterization and infection of the primaryepithelial cells, we observed that EBV DNA was routinely detectablein the cultures prior to infection (data not shown). Since weemployed highly sensitive PCR to detect the viral DNA, we couldnot distinguish whether the signal was derived from infectedcells or from virions, virus fragments, or cell debris frominfected cells released by the tonsil explants. To distinguishbetween these possibilities, we performed RT-PCR analysis fora range of viral genes, including EBNA1U-K, EBNA1Q-K, EBNA2,LMP1, and LMP2 genes. To our surprise, RNA for LMP1 and onlyLMP1 was robustly and routinely detected in the cultures from21 of 30 EBV-positive tonsils but in 0 out of 15 cultures fromEBV-negative tonsils. Limiting dilutions of Southern-blottedLMP1 RT-PCR products (Fig. 5) illustrate that transcript levelsin 2 x 10⁵ to 1 x 10⁶ epithelial cells were equivalent to 10to 100 control EBV-infected IB4 B cells. By comparison, signalsfor the other genes were detected in only one or two culturesat or close to the detection limit, casting doubt on the significanceof these findings. LMP1 is expressed both in latently infectedcells and in cells replicating the virus. However, the replicativeform is truncated and lacks exon 1. The PCR primers used weredesigned to bridge exons 1 and 2 and therefore detect only thelatent form of LMP1 (53). The reproducible detection of thelatent form of LMP1 mRNA demonstrated that the primary culturescontained latently infected cells.

View larger version (28K):
[in this window]
[in a new window]
FIG. 5. RT-PCR for LMP1 expression in primary ex vivo epithelial cells from the palatine tonsil. RT-PCR was performed on RNA from 2 x 10⁵ to 1 x 10⁶ day 10 to 14 epithelial cells. The PCR products were fractionated on agarose gels, Southern blotted, and hybridized with an LMP1-specific probe. The size of the PCR product (150 bp) is indicated. Limiting dilution was performed with serially diluted EBV-positive IB4 cells as a positive control (+). EBV-negative BJAB cells and H₂O were negative controls (–). Epithelial cultures of 15 individual tonsils are shown; 12 were from EBV-positive donors (T), 9 of which were positive for LMP1, and 3 were from EBV-negative donors (T*), none of which were positive for LMP1.

It is unlikely that the LMP1 transcript signals we detectedwere from viable infected B cells because the cultures werewashed frequently with medium to remove nonadherent cells andthe epithelial cells were split twice before analysis by gentlywashing them with trypsin prior to transfer to 100-mm-diameterdishes. By that time, no B cells could be detected in the culturesby histochemistry or flow cytometry. Moreover, the tissue cultureconditions were highly selective against lymphocytes and othernonepithelial cells. Control experiments with primary B cellsor EBV-transformed lymphoblastoid cell lines showed that allthe B cells were dead after 2 to 3 days of culture in the selectivemedium (data not shown). This raised the possibility that therewere latently infected epithelial cells in the cultures. Toidentify which type of cell was infected, we performed immunochemistryand fluorescence staining to detect LMP1 protein expression(Fig. 6). We chose LMP1 because the gene was the only one thatwe found to be reproducibly expressed at the transcript levelin the cultures. In control experiments, we were able to reliablydetect a single EBV-positive LMP1-expressing cell in a backgroundof 10⁵ EBV-negative cells, indicating the sensitivity of thisstaining (data not shown). In double-blind studies, we detectedLMP1 protein expression in single cells with clear epithelial-cellmorphology from 5 of 19 independent 7- to 10-day-old primaryepithelial cultures. Subsequent analysis revealed that all fivealso expressed LMP1 RNA, based on RT-PCR for latent LMP1 transcripts,and were derived from EBV-positive tonsils, based on DNA PCR(Fig. 6C to H), whereas 10 LMP1-negative cultures were derivedfrom EBV-negative tonsils (Fig. 6B). Thus, LMP1-expressing epithelialcells were found in cultures from 5 of 9 EBV-positive tonsilsand 0 of 10 EBV-negative tonsils. The level of infection basedon LMP1 protein staining was $~$ 5 to 10 per 10⁵ epithelial cells(0.005 to 0.01%), quite similar to the levels of infection reportedfor the B-cell compartment (27). LMP1 (latent) transcripts weredetected in all the cultures where LMP1 protein was found. Wewere unable to find evidence of EBNA1, EBNA2, or LMP2 expressionin any epithelial cells. The absence of EBNA1 may imply thatthe infected cells were not proliferating (12), and this wouldbe consistent with the observation that the LMP1-positive cellswere always found alone, never in clusters. However, as withall histochemical and immunofluorescence studies, it is impossibleto resolve whether the failure to find a protein representsabsence of the protein or whether the levels of expression weretoo low to be detected.

View larger version (75K):
[in this window]
[in a new window]
FIG. 6. LMP1 protein expression in ex vivo tonsillar epithelial cells. Epithelial cells were cultured as described in Materials and Methods and stained between days 7 and 10 with the anti-LMP1 monoclonal (S12) primary antibody for HC or IF. (A) LMP1 expression was assessed in the EBV-positive control B-cell line IB4. (B) Epithelial-cell culture from an EBV-negative tonsil. (C and D) Epithelial-cell cultures from EBV-positive tonsils. Left panel, low power; right panel, high power. (E and F) Epithelial-cell cultures from EBV-positive tonsils analyzed by HC (E) or IF (F). (G and H) Epithelial-cell cultures from EBV-negative tonsils infected in vitro and analyzed by HC (G) or IF (H). Note the similarity in staining observed when LMP1-expressing epithelial cells from an EBV-positive tonsil (ex vivo [E and F]) are compared with epithelial cells from an EBV-negative donor infected in vitro (in vitro infection [G and H]). Panel F shows a typical epithelial perinuclear distribution of LMP1 in an ex vivo-infected cell.

One possible explanation for the presence of these infectedepithelial cells comes from the well-known phenomenon that Bcells latently infected with EBV will replicate the virus whenplaced in culture. It was conceivable, therefore, that the epithelialcells became infected through cell-to-cell contact with suchcells. This interpretation is unlikely, because the explantcultures were continuously maintained in the presence of a highconcentration of acyclovir (0.1 mM) in the media. Acycloviris a common antiviral agent (29) that effectively blocks viralreplication with a 50% effective dose of 0.3 µM, basedon blocking viral replication in a lymphoblastoid cell linein vitro. We conclude, therefore, that there are bona fide LMP1-positive,latently infected epithelial cells in the cultures.

Viral replication is a frequently reported consequence of EBVinfection of epithelial cells in culture which might be differentiationdependent (10, 22, 23, 28, 47, 54). Since our cultures wereroutinely grown in the presence of acyclovir, it would not bepossible to detect the presence of viral replication. Therefore,we derived epithelial cell cultures in which acyclovir was removedafter day 4. The cells were further incubated for 4 to 6 daysin the absence of acyclovir in selective media. The epithelialcells were then stained with a cocktail of antibodies recognizingimmediate-early (BZLF1), early (EAD), and late (VCA) viral lyticproteins. Lytic antigen expression was detected in two of twocultures from EBV-positive tonsils but in zero of two culturesfrom EBV-negative tonsils (Fig. 7A to I). Most of the lyticstaining was nuclear, indicating EAD or Z protein expression,but occasional cytoplasmic staining (VCA) was also observed(Fig. 7B and C). The cells expressing lytic antigens were usuallyfound in clusters, suggesting that the virus may have spreadin the cultures through cell-to-cell contact between epithelialcells. This is a different pattern from what we observed withLMP1, which was always detected only in single cells. This indicatesthat different cells were expressing LMP1 and the lytic antigens.These data show that cells in primary tonsil epithelial culturesfrom healthy EBV-positive individuals probably sustain bothlatent and lytic EBV infection.

View larger version (43K):
[in this window]
[in a new window]
FIG. 7. (A to I) Immunofluorescence images of lytic gene expression in primary cultures from two tonsils. Cells were cultured in the presence of acyclovir until day 4. On day 5, SFM plus KGF was added, and the cells were transferred on days 7 and 8 to culture slides. Cytospins of 8- to 16-h IgG-treated Akata cells were used as positive controls (J to M). VCA staining is cytoplasmic (J), whereas Zebra (Z) and early antigen (Ead) staining is nuclear (K and L). Epithelial cells were fixed and stained on day 10 with the pooled monoclonal antibodies against Ead, Z, and VCA. The antibodies were pooled to increase sensitivity (M). The white arrows in panels B and C point to cytoplasmic VCA staining. Merged images are presented to emphasize the predominantly nuclear nature of the positive staining in both cultures derived from Z and Ead (D to L).

DISCUSSION

Top

Discussion

The results presented in this paper relate to three importantareas of EBV interaction with epithelial cells. These are theestablishment of a representative primary culture system fora hypothesized target of infection in vivo (tonsil epithelium),the demonstration that a small subset of such cells is infectiblein vitro by exogenous virus via cell-cell contact, and the observationthat, without adding any exogenous virus, EBV-infected epithelialcells are already present in cultures from EBV-positive individuals.

Tonsil epithelium is a heterogeneous lymphoid organ containingtwo different types of actively differentiating epithelia, thelining stratified squamous epithelium and the reticulated crypt,or lymphoepithelium. The tonsil crypts represent a specializedcompartment, important in the immunological functions of thetonsil because of the immediate proximity of the epithelialand lymphoid tissues (38). Human immunodeficiency virus hasbeen reported to specifically target this region, possibly becauseof the close interaction of human immunodeficiency virus targets(15). Keratin expression patterns confirmed that our proceduregenerated heterogeneous epithelial-cell cultures. K8, K18, andK19 expression in subsets of cells suggests that they were derivedfrom the reticular crypt epithelium, and the appearance of largedifferentiated cells was consistent with an ongoing active processof differentiation in the cultures (11, 32). It is possiblethat the patterns of keratin expression we observed were inducedin vitro. However, primary oral epithelial cells possess intrinsicallydetermined programs for keratin expression, which are believedto be unaffected by short-term in vitro culture (30). Interestingly,early removal of the tissue explants abolished the growth andviability of the remaining adherent epithelial cells. It isconceivable that culturing epithelial cells from tissue explantsprovides a desirable site-specific mesenchymal influence ininitial stages of culture. Mesenchymal cells are absent fromcultures in which epithelial cells are isolated from enzymaticallydigested tissues. This culture procedure often requires an artificialfeeder layer, which might benefit the growth of particular subpopulationsof epithelial cells (24, 31). Our culture dynamics support thenotion that primary epithelial cells are sensitive to theirmilieu (i.e., stroma), which is likely responsible for the contextualspecificity of various epithelia (37, 40).

We consistently detected EBV DNA, RNA, and proteins in the cultures.The question arises as to the origin of these infected cells.The obvious explanation is that we were detecting contaminatinginfected B lymphocytes. There are several lines of evidencethat eliminate this possibility. The level of infected B cellsin the tonsil is $~$ 1 in 10⁵ (27), so to have a single infectedcell would require the presence of $~$ 10⁵ uninfected lymphocytes,which we would readily detect in the cultures. Furthermore,the selective medium used in the cultures kills all lymphocytes,including EBV-infected B cells, within a few days. Most compelling,though, was the fact that the cells staining positive for LMP1and lytic proteins had the typical morphology of epithelialcells. Indeed, the elimination of all the lymphocytes, includingEBV-infected B cells, may have been important in increasingour chances of detecting the rare infected epithelial cells.Where and how were these cells infected? The only possibilityis through contact with infected B cells in the explants, sincethe epithelial cells cannot be infected by cell-free virus.Infected B cells in the explant would have survived for onlya day or two in the selective medium. Therefore, the epithelialcells would have to have become infected before or shortly afterthey began migrating onto the tissue culture dish. Since thisoccurred in the presence of acyclovir, we must conclude thatthese B cells were already producing infectious virus at thetime of explantation. This in turn leads to the inevitable conclusionthat the epithelial cells were infected by virus-producing Bcells that were already replicating EBV in the tonsil. Alternatively,the epithelial cells were already infected at the time of explantation.It is unlikely that we will ever be able to distinguish betweenthese two possibilities, because the number of infected epithelialcells is so small that it would be impossible to convincinglydetect them on a consistent basis in the presence of all ofthe lymphocytes present when staining cross sections of intacttonsil tissue. Mechanistically, however, the two explanationsboth lead to the same conclusion, namely, that normal tonsilepithelium is infectible with EBV in situ (whether it occurredin vivo or in vitro), and support the proposed dual tissue tropismof EBV for B and epithelial cells of the crypt regions of thetonsil (8, 37, 59).

The presence in our cultures of both latently and lyticallyinfected cells is fully consistent with their expected propertiesbased on the behavior of EBV in the major associated epithelialpathologies, NPC and OHL, and other previous reports of EBV-infectedepithelial cells in vitro (10, 22, 47). Whether our experimentsare indicative of a role that epithelium may play in EBV persistenceremains unresolved. It might be important to note that the frequencyof EBV-infected cells in the primary cultures was similar tothat found in B cells from the same tonsils ( $~$ 0.01%). An averagetonsil has a total epithelial surface area of 340 cm² (37).This would represent $~$ 5 x 10⁷ tonsil epithelial cells in vivo(based on the confluency of epithelial cells in a 100-mm-diameterdish), which translates into $~$ 5,000 infected epithelial cellsper tonsil at any given time. This is comparable to the estimatesof the total number of infected B cells in the tonsils, whichis $~$ 10,000 to 50,000 (27).

If epithelial cells do play a role in viral persistence, thefailure to infect them with cell-free virus suggests they arenot the target of primary infection. Their ability to be infectedthrough cell-to-cell contact means that they could mediate therelease of EBV into saliva from infected B cells in the lymphoepithelium.The ready ability of epithelial cells to replicate EBV, combinedwith the observations that persistent infection is not sustainablein the absence of B cells (17), further supports the idea thatthe primary role of epithelial cells is to facilitate the releaseof EBV, not to maintain persistent infection, which is presumablythe function of the B-cell compartment. What, then, is the roleof the latent infection we have observed? If epithelial cellscannot sustain a persistent infection, we must conclude thatlatent infection is transient and either an epiphenomenon ofthe ability of epithelial cells to be infected or possibly amechanism for short-term persistence, which might provide amore stable source of cells producing infectious virus.

Virion infection of polarized epithelial cells has been reported,but only through the basolateral surface (54). It will be necessaryto discover conditions under which our primary tonsillar epithelialcells can be infected at the apical (mucosal) surface with cell-freevirus before we can speculate that the epithelial-cell layermay also have a role as the first target tissue in primary EBVinfection.

There have been an ever-growing number of reports of successfulinfection of epithelial cells in vitro. Usually, these studieshave employed established cell lines (10, 14, 16, 22, 23, 28,45-47, 49, 54, 60). Attempts to infect primary nasopharyngeal,ectocervical, and gastric epithelia have been described, withvariable success. Nasopharyngeal tissue was reported to be uninfectible(16); the ectocervix was reported to be infectible only by oropharyngealstrains of EBV, not laboratory isolates (46); and the gastricepithelium was reported to be infectible in only 3 of 21 attempts(36). We believe, therefore, that this represents the firstreport of a technique to reproducibly infect representativeprimary cultures from tonsil epithelium. Reports of viral-geneexpression patterns in infected epithelium have also varied.For example, EBV-infected lung, colon, and primary gastric epithelialcells probably do not express LMP1 upon EBV infection (10, 36,46), but in other reports, EBV-infected epithelial cells ofnaso- and oropharyngeal origin do express LMP1 (10, 54). Thesediscrepancies may in part reflect technical issues but may alsoarise because epithelia from different sites are heterogeneousand biologically distinct. We think, therefore, that it is importantto study EBV infection in the correct epithelial-tissue background.This concept is reinforced by studies of EBV in B cells, whereit is clear that the pattern of viral genes expressed is dependentupon the type and location of the infected cell (5). It wasstriking that the only latent gene we found reproducibly inthe ex vivo-infected cells was the viral oncogene LMP1. At thispoint, it would be premature to suggest that this reflects anovel form of EBV gene expression, i.e., LMP1 only. This isbecause the lower limit of the detection sensitivity for theRT-PCR and immunohistochemistry assays for the other viral latentgenes is not known and the infected cells are so rare. Thismakes a negative result difficult to interpret with confidence.Nevertheless, the presence of LMP1 suggests that it is automaticallyexpressed upon EBV infection of normal nasopharyngeal epithelium.Such expression of the oncogenic LMP1 could be the initiatingevent that may ultimately lead to development of nasopharyngealcarcinomas.

In vitro infection of our primary tonsil epithelial culturesrequired cell-cell contact, and only a small subset of cellswas susceptible to infection. These observations are consistentwith the findings of others using different epithelial-celltypes (10, 23, 54). There is increasing evidence that virusesexhibit a high degree of specificity for their targets (7).It would be interesting to see what specific characteristicsthe infectible cells might have and if expression of particularviral-gene transcription programs relates to the differentiationstage of the cells. Evidence for infection specificity in epithelialcells has recently been shown in vitro and might depend on theexpression of ß1 integrin (54), which is likely tocorrelate with the differentiation stage (54, 57). Our primaryin vitro model with actively differentiating cells might provea valuable tool to address these issues in the future.

Early findings of EBV in oropharyngeal epithelial cells fueledthe notion that epithelium might play a role in viral persistence(1, 45-47). However, subsequent reports were unable to reproducethese findings (25, 35), casting doubt on the role of epithelialcells. Recent findings in culture defining an epithelial-cellreceptor (8, 54) for the virus and suggesting a mechanism toexplain the potential dual tissue tropism of the virus for Bcells and epithelial cells (8, 54) have reopened the issue.However, what is lacking is convincing evidence of infectionin vivo. It has not been possible to detect EBV-infected epithelialcells in vivo using EBER hybridization in combination with B-celland epithelial-cell-specific markers on cross sections of tonsilsand in exfoliated epithelial cells in throat washes of infectious-mononucleosispatients (25, 35). This approach was based on the assumptionthat during infectious mononucleosis there should be a massiveamount of epithelial-cell infection that matches the levelsseen in B cells, where as many as 50% of the memory cells maybe infected (21). From in vitro studies, it is well known thatmost, if not all, B cells are infectible with EBV; however,the in vitro experiments shown here and by others (54) suggestthat only a small fraction of tonsil epithelial cells are infectiblewith EBV. Therefore, massive infection of epithelial cells maynot be possible in vivo during infectious mononucleosis because,unlike B cells, most epithelial cells are not targets of infection.This could explain the failure to detect infected epithelialcells in the tonsils of acute infectious mononucleosis patientsby using histochemical techniques.

Our findings support the developing model in which EBV is ableto infect healthy epithelium as part of its life cycle (8, 39,54, 55) and contribute to our understanding of the evolutionof naso- and oropharyngeal epithelial pathologies.

ACKNOWLEDGMENTS

We thank Jim Rheinwald, Harvard Skin Disease Research Center,Brigham and Women's Hospital, for advice on the culture andcharacterization of human tonsil epithelial cells and CherylGreen from the Massachusetts General Hospital, Boston, for providingthe tonsils.

This work was supported by Public Health Service Grants AI 18757and CA 65883.

FOOTNOTES

* Corresponding author. Mailing address: Department of Pathology, Jaharis Bldg., Tufts University School of Medicine, 150 Harrison Ave., Boston, MA 02111. Phone: (617) 636-2726. Fax: (617) 636-2990. E-mail: david.thorley-lawson{at}tufts.edu.

REFERENCES

Top