|Home | About | Journals | Submit | Contact Us | Français|
The current model of Epstein-Barr virus (EBV) infection and persistence in vivo proposes that EBV uses the germinal center (the GC model) to establish a quiescent latent infection in otherwise-normal memory B cells. However, the evidence linking EBV-infected cells and the GC is only indirect and limited. Therefore, a key portion of the model, that EBV-infected cells physically reside and participate in GCs, has yet to be verified. Furthermore, recent experiments suggested that upon infection of GC cells the viral growth latency transcription program is dominant and GC functionality and phenotype are ablated, i.e., EBV infection is not consistent with GC function. In this study we show that in vivo, EBV-infected B cells in the tonsils retain expression of functional and phenotypic markers of GC cells, including bcl-6 and AID. Furthermore, these cells are physically located in the GC and express a restricted form of latency, the default latency program. Thus, the EBV default latency transcription program, unlike the growth latency program, is consistent with the retention of GC functionality in vivo. This work verifies key components of the GC model of EBV persistence and suggests that EBV and the GC can interact to produce the latently infected memory cells found in the periphery. Furthermore, it identifies latently infected GC B cells as a potential pathogenic nexus for the development of the EBV-positive, GC-associated lymphomas Hodgkin's disease and Burkitt's lymphoma.
Epstein-Barr virus (EBV) is able to promiscuously and efficiently infect any resting B cell in culture, leading to cellular activation, proliferation, and the outgrowth of transformed lymphoblastoid cell lines (the growth latency transcription program, also known as latency 3) (38, 48). Not surprisingly, the virus is also associated with important human malignancies that include B-cell lymphomas (Burkitt's, Hodgkin's, and immunoblastic lymphomas in the immunosuppressed) and carcinomas (nasopharyngeal and gastric) (38, 50). Despite its pathogenic potential, EBV benignly infects >90% of the human population and persists for life. A central question in understanding both persistent infection and the origins of EBV-associated cancer is how the host and virus interact to allow benign persistent infection. Specifically, how does this interaction regulate and limit the capacity of the virus to drive cellular growth?
An important insight into this question was the discovery that EBV persists in circulating, resting, memory B cells (5, 50) These cells do not express viral proteins (17) and appear to be maintained in the periphery by normal memory B-cell homeostasis mechanisms, not by the growth-promoting activity of the virus (44). Consequently, they are not a pathogenic risk to the host nor are they subject to immunosurveillance and can therefore persist benignly within the human host for decades.
The mechanism by which EBV gains access to the memory B cell remains controversial. The most widely held model is that the virus uses the growth latency program to drive newly infected resting B cells into the cell cycle so that they can differentiate into the resting memory state via the germinal center (GC) reaction, the GC model (50). The GC is the region of secondary lymphoid tissue where antigen-activated B cells undergo proliferation, class switch recombination (CSR), somatic hypermutation (SHM), antigen selection, and affinity maturation (28, 29, 31). In this highly competitive environment, failure to successfully compete for antigen and T-cell help leads to apoptosis, whereas success results in the production of plasma cells or memory cells (29). Because GC B cells are hyperproliferative and undergoing active processes that involve double-stranded DNA breaks (CSR) and mutation (SHM), they are considered at high risk for genetic abnormalities leading to B-cell lymphoma (22, 37, 42). Both CSR and SHM are mediated by the enzyme AID (activation-induced cytidine deaminase) (34), which is highly expressed in GCs. AID has recently been shown in a mouse transgenic model to be required for the generation of the immunoglobulin (Ig)/c-myc translocations characteristic of the EBV-associated tumor Burkitt's lymphoma (12). Thus, the potential intersection of EBV and the GC reaction, both of which drive B-cell growth and which, respectively, also provide antiapoptotic signals and promote genetic instability, provides a powerful combinatorial risk factor for development of lymphoma.
One limitation of the GC model is that the evidence linking EBV and the GC is only indirect and limited. We have shown previously that the tonsils of individuals persistently infected with EBV contain infected cells bearing the GC marker CD10 (6). Less than 0.2% of the cells were undergoing lytic replication (27); rather, the cells were latently infected, expressing a limited set of latent genes which included LMP1, LMP2, and EBNA1Q-K (the default latency program) (50). The same viral transcription program is found in Hodgkin's disease (38), a tumor of the GC (46). LMP1 and LMP2 have been shown to possess, respectively, the CD40 and BCR signaling functions necessary for GC survival (7, 8, 13, 15, 35, 47), and EBNA1 is required for replication of the viral genome. However, CD10 is only a phenotypic marker of GC cells. It has no known GC-specific function and is also expressed on immature B cells. Furthermore, it is known that EBV latent proteins like LMP1 can profoundly affect the surface phenotype of the cells it infects. Therefore, the presence of CD10 on latently infected tonsil B cells is not definitive of a GC origin and could be an artifact of virus infection. Immunohistochemical analysis has demonstrated the presence of rare, usually single, EBV-infected cells in the GCs of healthy carriers of the virus (20). However, the nature of these cells and the viral genes they express were not explored and we now know that GCs are highly dynamic structures that are often traversed by other B cells (40). Therefore, these cells could be latently infected memory cells temporarily entering the GC. In sum, a key portion of the GC model, that EBV-infected cells physically reside and participate in GCs, has yet to be verified.
The GC model has been further challenged based on studies demonstrating that upon in vitro infection of GC cells with EBV the lymphoblastoid phenotype, induced by the growth latency program, is dominant and the infected cells lose their functional and phenotypic GC markers (43). This is consistent with earlier findings that the dominant population of EBV-infected cells, identified by microdissection and present in the GCs of tonsils from individuals acutely infected with EBV, were undergoing clonal expansion driven by the viral growth latency program and did not express GC markers (26). Together these studies imply that latent infection with EBV and GC functionality are not compatible and have led to an alternate model of EBV persistence. In this model EBV-infected cells do not undergo a GC reaction but instead the virus directly infects memory and/or GC B cells that initially undergo clonal expansion driven by the viral growth latency program and then through some unspecified selection process enter into a resting state (25, 26). We argued in a previous study (14) that persistent EBV infection is already established at or near long-term levels by the time patients arrive in the clinic, and therefore the number of EBV-infected cells undergoing a GC reaction would be too small to be detected by traditional microdissection techniques. Furthermore, we had shown earlier (6) that direct infection of GC cells in vitro gives rise to the growth latency transcription program, not the default latency program, and argued therefore that CD10+ cells from tonsils that express the default latency program must arise from another mechanism than direct infection with the virus. We further argued that what was being observed in the GCs of acutely infected individuals was a consequence of direct infection with EBV which, just like in vitro, always leads to the dominant lymphoblastoid phenotype. This type of infection would be transient as a consequence of early events occurring prior to the initiation of the immune response and would be eliminated by cytotoxic lymphocytes (CTL), leaving only small numbers of EBV-infected cells truly undergoing a GC reaction in long-term persistent infection.
Thus, the issue of whether EBV-infected cells expressing the default latency transcription program physically enter into and participate in the GC, which is central to our understanding of EBV persistence and lymphomagenesis, remains to be settled. To address this, we have asked if the EBV-infected CD10+ tonsil cells expressing the viral default latency program, which we have described previously, express the classical functional and phenotypic markers of GC cells and whether they are physically located in GCs.
Human palatine tonsils were obtained from patients 18 years or younger receiving a tonsillectomy performed at the Tufts New England Medical Center, Boston, MA. Tonsil tissues were minced in phosphate-buffered saline (PBS) plus 10% bovine serum albumin (BSA), and mononuclear cells were isolated by using Ficoll-Paque Plus (Fisher) centrifugation as described previously (6). This study was approved by the Tufts Medical Center Institutional Review Board.
The EBV-positive lymphoblastoid cell line IB4 (gift of Eliot Kieff) was used as a positive control for DNA PCR of the W-repeat region of the EBV genomes, and TaqMan reverse transcription-PCR (RT-PCR) was used for EBER1, LMP1, and LMP2. The EBV-positive Burkitt's lymphoma cell line Rael (gift of Sam Speck) was used as the positive control for EBNA-1 Q-K TaqMan RT-PCR. The EBV-negative cell line CB59, a mouse T-cell hybridoma cell line (gift of Miguel Stadecker) was used as a negative control in all experiments. All cell lines were cultured at 37°C with 5% CO2 in RPMI 1640 supplemented with 10% fetal bovine serum, 2 mM glutamine, 2 mM sodium pyruvate, and 100 IU of penicillin-streptomycin (RPMI complete).
Tonsil B-cell subpopulations were purified by fluorescence-activated cell sorting (FACS). Cells were resuspended to 5 × 106 cells/100 μl in PBS-5%BSA and placed in Eppendorf tubes. For cell surface labeling only, the appropriate concentration of fluorochrome-conjugated antibody was added to the tube and incubated for 15 min at 4°C and the cells were then washed. For intracellular staining the fluorochrome-conjugated antibodies against cell surface markers were added to cells and incubated for 15 min at room temperature. Cells were washed, fixation solution (1% paraformaldehyde in PBS) was added, and incubation continued for 15 min at room temperature. Then, cells were washed, permeabilization solution (0.5% saponin in PBS) was added, and cultures were incubated for 20 min at room temperature. Fluorochrome-conjugated antibody to the intracellular marker was added, incubated for 15 min at room temperature, and washed. For indirect stains an additional step of incubations and washes was included. Cells were analyzed and sorted by the MoFLO FACS system at the Tufts University Laser Cytometry Unit. Each sorted population underwent reanalysis to ensure >90% purity. For a list of markers tested, see Table Table1,1, and for antibody source, dilution, and fluorochrome information see Table S1 in the supplemental material.
Limiting dilution analysis was used to measure the frequency of EBV-infected cells by performing DNA PCR for the W-repeat region of the EBV genome on replicates of serially diluted cells and detection of the PCR products by real-time PCR as described in reference 14. The frequency of cells expressing a particular gene, EBER1, EBNA1Q-K, LMP1, or LMP2, was also estimated by limiting dilution analysis and RT-PCR for each gene, as detailed in reference 18, except PCR products were detected by real-time PCR as described below. In all cases the frequency of cells was estimated using Poisson statistics.
Cuboidal tonsil tissue pieces were submerged in RNALater (Ambion), placed in histology blocks with OCT compound, and flash frozen in liquid nitrogen. Tissue histoblocks were sent to the NEMC Pathology Lab, where they were sectioned (10 μm), three slices per slide, using a −30°C microtome and stored at −80°C. Slides were washed with 70% ethanol in distilled RNase and DNase-free water on ice for 2 h, stained with 20% hematoxylin solution in RNALater for 10 s, washed, then stained with 10% eosin solution in RNALater, and dried with a series of ethanol and xylene washes. Stained sections were examined under a microscope and dissected by scraping out the cells using a sterile scalpel and placing in an RNase-free Eppendorf tube filled with PBS plus 5% BSA on ice. GCs were pooled from 1 to 30 slides. Each slide contained three tissue slices with 10 GCs per slice. For the negative control dissections, broad swaths of the EBV-negative tonsil tissues were taken that included multiple GCs, mantle zones, and interfollicular cells.
RNA was purified by TRIzol extraction (Invitrogen) and then treated with DNase I (Invitrogen) to eliminate DNA prior to RNA amplification. cDNA was synthesized from RNA using an Invitrogen cDNA kit. For the cDNA synthesis reaction mixture a master mix was prepared which included 4 μl of 5× iScript reaction mix, 1 μl of iScript reverse transcriptase, and 8 μl of nuclease-free water. Seven μl of purified RNA was added to 13 μl of master mix. All reactions were performed on an Applied Biosystems PCR machine. The protocol was as follows: one cycle that included 5 min at 25°C, 30 min at 42°C, and 5 min at 85°C. For the RT-PCR a master mix was prepared, containing 12.5 μl of IQ Supermix (Bio-Rad), 2.5 μl of 900 nM primers, and 2.5 μl of 250 nM fluorogenic probe. Five μl of cDNA was added to 20 μl of master mix with a final reaction volume of 25 μl. All RT-PCRs were performed on a Bio-Rad iCycler. The protocol was as follows: step 1, one cycle of 3 min at 95°C; step 2, 55 cycles of 15 s at 95°C and 1 min at 60°C. All RT-PCR assays were optimized to detect down to the single cell level. See Table S2 in the supplemental material for a list of primers and probes.
For each tonsil to be analyzed by scraping/dissection a piece was set aside for isolation of the GC cells (CD19+ CD10+) by flow cytometry. A series of dilutions of 105, 104, 103, and 102 of these GC cells was then made and aliquoted into sterile Eppendorf tubes, with five replicates for each dilution. Each of the replicates was further diluted 1/2, 1/4, and 1/8. cDNA was synthesized and RT-PCR for CD10 was then performed on each dilution. A calibration curve of cell number versus CD10 RT-PCR cycle number was plotted and used to convert the CD10 RT-PCR signal obtained from the pooled dissected GC cells, from the same tonsil, into an estimate of cell number.
In a series of control experiments we observed that the EBER1 copy number in CD10+ GC cells isolated by flow cytometry was very similar to the copy number in the EBV+ lymphoblastoid cell line IB4. Therefore, we generated an EBER1 RT-PCR calibration curve by performing EBER1-specific RT-PCR on cells from the IB4 EBV lymphoblastoid cell line serially diluted with cells from an EBV-negative tonsil. A calibration curve of cell number versus EBER1 RT-PCR cycle number was plotted and used to convert the EBER1 RT-PCR signal obtained from the scraped/dissected GC cells into an estimate of infected cell number.
The expression of AID and bcl-6 are defining functional features of germinal center B cells. AID is an enzyme of the APOBEC family that is highly expressed in GCs (22, 33, 34). It is the enzyme necessary to initiate SHM and CSR, the functional sine qua non of the GC. bcl-6 on the other hand is the master transcription factor of the GC (11, 52). Its expression is restricted to GC cells, it's required for GC production, and its downregulation is essential for B cells to leave the GC (2) (see Table Table11 for a list of all the markers used in this study and their GC-related functions). If CD10+ EBV-infected cells are undergoing a GC reaction, then they should express both of these molecules. To test this, tonsil B cells from a healthy carrier of EBV were stained for CD10 and counterstained for either bcl-6 or AID. Since there are few, if any, CD10+ bcl-6− or CD10+ AID− cells we compared the frequency of EBV-infected cells in the CD10+ population before and after counterselection for the second GC marker. If the EBV-infected cells were true GC cells then they should copurify with the double-positive population. If on the other hand the EBV-infected cells were bcl-6 and/or AID negative then EBV-infected cells would be depleted or completely absent from the double-stained population (Fig. (Fig.1).1). As a control for sensitivity of the assay we also isolated the IgD+ population of tonsil B cells, which we have shown previously to have a significantly lower frequency of infected cells than the CD10+ population (6). The frequency of virus-infected cells was estimated using a limiting dilution assay, which we have described previously (14, 32), where the infected cells are detected by DNA PCR for the W-repeat sequence, and which has a maximum error of approximately ±30%(21). The results are summarized in Table Table2.2. For three separate tonsils we found that the frequency of EBV-infected cells in the CD10+ population did not change significantly upon counterselection with either bcl-6 or AID. In comparison the frequency of infected cells in the IgD+ population was consistently lower, as expected, confirming that our technique could reproducibly distinguish significant reductions in the frequency of EBV-infected cells. As a negative control we performed the same analysis for c-myc expression. c-myc is a major regulator of cell growth and apoptosis in B cells that is reportedly not expressed in the GC (22, 23, 30, 41, 42). In this case if the EBV-positive CD10+ cells were true GC cells then they should colocalize with the CD10+ c-myc− population. As shown in Table Table22 this was indeed the case for three separate tonsils. We conclude therefore that the CD10+ cells latently infected with EBV specifically express functional markers of the GC.
The localization of lymphocytes within lymph nodes is regulated via the response to chemokines, the receptors for which are differentially expressed by different subsets of cells. The expression of specific chemokine receptors is therefore another set of functional markers for the GC, which in addition indicate the localization of the cells. CXCR4 and CXCR5 are chemokine receptors that are highly expressed on GC cells (39) and are thought to be essential for proper localization of centroblasts and centrocytes, respectively (1). Using the same approach as described in the previous section, we analyzed the expression of these receptors on the CD10+ cells latently infected with EBV in the tonsils of healthy carriers of EBV. As shown in Table Table22 they expressed both CXCR4 and CXCR5. In this case as a negative control we used the chemokine receptor CCR7. CCR7 is expressed weakly on naive B cells but becomes upregulated upon antigen-driven activation, causing the cell to migrate into the follicle, where its expression is downregulated. Hence, it is not present on GC cells (39). We found that it was not expressed on the CD10+ EBV-infected cell population (Table (Table2).2). We conclude therefore that the CD10+ cells latently infected with EBV express the expected chemokine profile for GC cells and therefore are likely to be localized in the GC.
CD38 is expressed on GC cells and highly expressed on plasma cells, although its function on either population is unknown. Despite this, CD38 is routinely used as a standard marker for GC cells, both in flow cytometry and immunohistochemical studies (36). We therefore again used the same approach described in the previous sections to analyze CD38 expression on the CD10+ EBV-infected population of tonsil cells. As can be seen in Table Table22 these cells also coexpressed CD38.
In summary the EBV-infected CD10+ population in tonsils expresses GC-specific functional markers and homing receptors, and these cells are phenotypically indistinguishable from GC B cells.
The data in the previous section suggest that EBV latent gene expression does not detectably influence the behavior of the latently infected GC cells. A hallmark of the GC is the progression of the cells through stages of proliferation (centroblasts) and resting (centrocytes). Centroblasts (CBs) are actively undergoing CSR and SHM and have downregulated the surface expression of their immunoglobulin (sIg) genes. When they undergo growth arrest to become centrocytes (CCs) they reexpress sIg and undergo antigen selection. Besides sIg and proliferation, CBs and CCs are routinely distinguished based on expression of the marker CD77 (positive and negative, respectively) (29), although its specificity has recently been challenged (19). CD77 is a neutral glycosphingolipid and like CD10 and CD38 has no known GC-related function.
To further analyze the behavior of EBV in the GC process we have tested for its presence in CBs and CCs fractionated based on sIg and CD77. The cells in each population were assessed for the expression of viral genes by using a limiting dilution RT-PCR that we have described previously (17) (see Materials and Methods). The genes tested were EBER1 (expressed in all infected cells), to estimate the frequency of infected cells, and LMP1, LMP2, and EBNA1Q-K (characteristic genes of the default latency program that we have previously shown to be expressed by the CD10+ EBV-infected population ). As can be seen in Table Table3,3, in all cases we found that EBV was equally distributed between the CB and CC populations and that both populations expressed the default latency transcription program. We conclude therefore that there was no discernible trend to suggest that EBV is preferentially retained in either population or that there was any evidence of differential latent gene expression between the two populations.
The studies presented above suggest that EBV-infected cells in the tonsil are undergoing a GC reaction and the chemokine receptor profile they express should cause them to migrate into and remain in the GC. To test directly if these cells physically reside in GCs we have designed a method to microdissect GCs from human tonsils and look for the presence of EBV. Sections of snap-frozen tonsils were placed on slides, stained with hematoxylin and eosin, and examined under a microscope to distinguish GCs from mantle zone and interfollicular regions (Fig. (Fig.2).2). Sections from multiple GCs were then scraped off the slide, pooled, and subjected to RT-PCR analysis for the GC-specific marker CD10. In parallel, CD10+ cells from the same tonsil were isolated by flow cytometry and used to generate a calibration curve of CD10+ cell number versus CD10 RT-PCR signal. The calibration curve and the CD10 RT-PCR signal from the scraped GC cells were then used to estimate the actual cell number in the scraped sample. Similarly, the RT-PCR signal for the EBV-specific small RNA EBER1 was used to estimate the number of EBV-infected cells in each pool (see Materials and Methods for details). In addition the pools were tested for expression of the genes characteristic of the default latency transcription program: LMP1, LMP2, and EBNA1Q-K. The collated results from three EBV-positive tonsils are shown in Table Table44 and indicate that EBV-infected cells are readily detectable in the pooled GCs and express the default latency transcription program. We believe these signals are specific and not caused by contamination for the following reasons. First, detection of EBV was dependent on the number of scraped cells in the samples. EBV gene expression was always detected when pools containing ≥104 cells were tested and never with fewer cells. This was consistent with, and expected based on, the frequency of EBV-infected cells we measured in the CD10+ population isolated by flow cytometry from the same tonsil, which typically was ~1 in 105 (see below). Second, detection of EBV by EBER1 RT-PCR always coincided with detection of the genes of the default latency program (32/32 samples), whereas default latency program genes were never detected when EBER1 was not detected (20/20 samples). Third, EBV-infected cells were not detected in control experiments where cells were scraped from tonsils previously determined to be EBV negative (samples labeled NCpool). Fourth, when cells were scraped and pooled from the mantle zones around the corresponding GCs, we detected no signals for any of the EBV genes we tested (not shown). We conclude therefore that the signals we have found cannot be due to fortuitous inclusion of mantle zone cells during the scraping process. Lastly, we were unable to detect an RT-PCR signal for CD10 in scraped cells from the mantle zone, interfollicular region, or epithelial/marginal zone region of the tonsil (not shown). This last control allowed two important conclusions. First, the sectioning technique did not smear cells across the tissue slice, and therefore our EBV-positive signals were actually located in the GC. Second, and most importantly, all of the CD10 signals detectable by our microdissection and RT-PCR studies reside within the physical demarcations of the GC. Since our CD10 RT-PCR techniques can detect the signal from the RNA equivalent to less than one CD10+ tonsil cell, separated by flow cytometry, we conclude that all detectable CD10+ cells reside physically within the GC. Therefore, the CD10+ EBV+ cells detected and analyzed by flow cytometry above must also reside physically within the GC.
We estimated the number of infected cells in the pooled GC samples based on the EBER1 RT-PCR signal and the total number of cells based on CD10 RT-PCR signal (see Materials and Methods). The results of this analysis are also included in Table Table4.4. As a positive control we included scraped cells from tonsils previously determined to be EBV negative, into which were spiked either 10 or 1 EBV-positive cell from the lymphoblastoid line IB4 (Table (Table1).1). We found that the estimated number of infected cells per pool ranged from 1 to 16. We could derive an estimate of the frequency of infected GC cells by taking the total estimated number of infected cells and dividing by the total estimated number of scraped GC cells. For the three tonsils these values were 7.4/105, 2.8/105, and 1.4/105, respectively (Table (Table5).5). We independently estimated the number of EBV-infected GC cells by performing DNA PCR for the W-repeat region of the EBV genome on limiting cell dilutions of CD10+ cells isolated from the same tonsils by flow cytometry. These values were, respectively, 1.2/105, 0.9/105, and 1.3/105 (Table (Table5),5), indicating that the two independent measurements were in good agreement despite the completely different techniques used (scraping and EBER1 RT-PCR versus flow cytometry and EBV W-repeat DNA PCR). This concordance of the two types of measurements validates our estimates of infected cells in the scraped pools. It also provides further proof that the signals in our pools were not derived through contamination, since it is highly unlikely that RNA/cDNA contamination in the RT-PCR would produce the same signals as detected by DNA PCR in cells isolated by a completely different method (flow cytometry).
We conclude that EBV-infected cells expressing CD10 and the viral default latency transcription program are physically located within the GCs.
In this paper we have verified a key tenet of the GC model of EBV persistence by showing for the first time that EBV-infected cells expressing the default latency transcription program are physically located in GCs and express functional markers, homing receptors, and the phenotypic markers of GC cells. These experiments further demonstrate that EBV can reside latently within GC cells in vivo without imposing the growth latency transcription program and the lymphoblastoid form of latency. What distinguishes infected GC cells in vivo is expression of the viral default latency transcription program which, unlike the growth latency program, appears to be fully consistent with retention of GC function. Since direct infection of GC cells in vitro (43) or in vivo (26) produces the growth latency program and obliterates the GC phenotype, it follows that the latently infected GC cells we have described in vivo cannot arise through direct infection and must be derived by another mechanism, i.e., differentiation of newly infected blasts (49). This work, taken together with previous evidence showing that EBV persists in GC-derived memory B cells (5, 44, 50) and that GCs are required for EBV to access these cells (9, 10), provides further evidence in support of the model that EBV uses the GC to access resting memory B cells as a site of persistence.
In virtually every aspect that we have examined, the EBV-infected GC cells in vivo are indistinguishable from antigen-activated GC cells and appear to be undergoing a normal GC reaction. This includes expression of the defining GC markers bcl-6 and AID. We may conclude therefore that the expression of CD10 on latently infected tonsil cells that we have described previously (6) is not an artifact due to the presence of the virus. Rather, it reflects the genuine status of these cells as GC cells. The cells also appear to be undergoing the process of cycling between centroblasts (sIglo CD77+) and centrocytes (sIghi CD77+). This is consistent with our ongoing view that EBV exploits rather than alters normal B-cell biology to establish and maintain persistent infection. It also explains our previous observations that the resulting latently infected memory B cells appear normal by every criterion we have tested (44, 45).
The one aspect in which the latently infected GC B cells do not appear to be normal is the absence of the massive clonal expansion characteristic of the GC. There are several possible explanations for this. The simplest is that the cells only undergo a few rounds of division before leaving the GC, perhaps because LMP1 and LMP2 signaling allows them to bypass the requirement for extensive expansion. The main objection to this is that mutational analysis of latently infected memory B cells in the periphery suggests that they have been through many rounds of proliferation in the GC (44). A second possibility is that the cells are indeed undergoing extensive proliferation but at the same time are being surveilled and killed by CTL. We favor this explanation because it is well-documented that there is a high frequency of activated CTL in tonsils that recognize latent proteins, including those expressed in the default latency program (16). A third possibility is that these represent latently infected memory cells that have migrated by chance into the GC and for some reason adopted a GC phenotype and presumably remained due to expression of GC-specific chemokine receptors. We think this is unlikely since there is no precedent for this type of behavior by memory B cells.
In our experiments viral gene expression is, of necessity, measured at the RNA level with the assumption that this correlates with protein expression. Given this caveat it appears that the GC process is not detectably altered by the expression of LMP1 and LMP2, which seem always to be coexpressed in the GC. What then is the role of LMP1 and LMP2 in the GC? Studies in vitro and from transgenic mice have imputed potent signaling activities to LMP1 and LMP2 (7, 8, 13, 15, 35, 47), implying that they can, in principle, drive the entire GC process in the absence of antigen and even rescue defective B cells (7, 47). However, this does not appear to be happening in vivo. The roles of LMP1 and LMP2 must presumably be more modest, perhaps only supplementing physiologic signals to provide a survival advantage in the highly competitive environment of the GC. For example, the EBV-infected cells may be antigen specific but would not have to compete effectively for antigen in order to survive, because LMP2 could supplement the necessary rescue signal. Similarly, the latently infected GC cells may only receive limited T-cell help, but LMP1 is capable of providing constitutive T-cell help signals. It has been suggested that extended T-cell help signaling preferentially pushes GC cells away from plasma cell differentiation and toward memory cells (3). Thus, the role of LMP1 may be to ensure that the latently infected GC cells enter the memory compartment, where they can persist, and not become plasma cells.
EBV and the GC are both potential sources of disease. EBV is a risk factor for cancer because of its growth-promoting activity. For example, unregulated LMP1 expression is oncogenic in transgenic mice (24), and LMP1 is constitutively expressed in Hodgkin's disease (4), implicating LMP1 in tumor development. EBV is also a potential risk factor for autoimmunity, because constitutive expression of LMP2 in transgenic mice can rescue autoreactive B cells (47), providing a possible mechanism for EBV in autoimmune diseases. Similarly, the GC reaction is characterized by SHM and CSR, which involve mutation and double-stranded DNA breaks that may favor tumor development and SHM of Ig genes, which could generate autoreactive B cells. Because of this, both EBV and the GC are tightly controlled and do not normally lead to pathogenesis; however, their interaction may greatly increase the risk of disease. For cancer this is clearly the case. The intersection of EBV and the GC provides a potential origin for Hodgkin's disease and Burkitt's lymphoma, both of which are associated with the GC and EBV. Whether the intersection of EBV and the GC also leads to autoimmune disease remains to be established.
Our results reveal the limitations of studies with cell lines and transgenic mice and the importance of corroborating these findings in human in vivo infections. For example, in vitro it has been shown that LMP1 downregulates bcl-6 (35) in B-cell lines, whereas it is apparent from our studies on GC cells in vivo that bcl-6 can be upregulated in the presence of LMP1. This led us to the conclusion that lymphoblastoid cell lines represent a potentially artifactual state that may not always reflect the normal biology of the virus in vivo. This conclusion presents itself because newly infected B cells in vivo do not appear normally to go through many rounds of cell division driven by the growth latency program and certainly have not been selected for efficient growth in tissue culture. Extensive proliferation of EBV-infected lymphoblasts may only arise in vivo in the context of pathogenic entities such as posttransplant lymphoproliferative disorder. It follows that observations made with lymphoblastoid cell lines should be substantiated in vivo and in some instances may better reflect on the pathogenic rather than the normal behavior of infected cells.
Similarly, problems arise in interpreting studies from transgenic mice, because in these studies single viral genes are expressed from nonviral promoters lacking the context of the whole virus (7, 8, 24, 47, 51). This produces inappropriate levels, locations, times, and contexts of expression. For example, in vivo, LMP1 is nearly always expressed in the context of LMP2, yet in transgenic mice the behavior of one LMP molecule is never studied in the context of the other. Furthermore, LMP1 is oncogenic (24) and LMP2 rescues autoreactive B cells (47) in transgenic mice, but neither happens in the healthy carrier state. Transgenic approaches are thus also revealed to have limited utility for studying the normal biology of EBV. Only in tumors is the expression of these potentially pathogenic latent genes constitutive due to other genetic lesions in the cells. These lesions presumably prevent the latently infected B cells from exiting the cell cycle, switching off latent gene expression, and becoming memory cells. Thus, transgenic experiments may be more informative for understanding the pathogenesis of EBV-associated diseases.
In conclusion, we have verified a key component of the GC model of EBV persistence and in doing so have shown an intimate interaction between the GC process and B cells latently infected with EBV. This interaction has the potential for explaining the origins of EBV-associated neoplastic and autoimmune diseases.
We thank Allen Parmelee and Steve Kwok for FACS.
This work was supported by Public Health Research grants CA65883, AI18757, and AI062989.
Published ahead of print on 4 February 2009.
†Supplemental material for this article may be found at http://jvi.asm.org/.