Differential Evolution and Stability of Epitope-Specific CD8⁺ T Cell Responses in EBV Infection¹

Epstein-Barr virus is a ubiquitous herpesvirus that infects more than 90% of the world’s population. During acute EBV infection, the virus replicates in lymphoid cells and oropharyngeal epithelium with the expression of the full array of viral proteins (1). The infection of B cells results in a life-long EBV carrier state in which B cells primarily express latent proteins (2). Whereas primary infection is usually asymptomatic in infants (3), adolescents and young adults frequently experience acute infectious mononucleosis (4). Studies spanning the past decade have shown that CD8⁺ T cell responses to EBV are important for controlling EBV replication (5, 6, 7, 8). Because individuals with acute EBV infection may be readily identified on the basis of the heterophile Ab test and the presence of atypical lymphocytes, EBV is an ideal model for the study of the generation and maintenance of CD8⁺ T cells in response to a persistent human viral infection.

The use of HLA class I tetramers in the lymphocytic choriomeningitis virus (LCMV)³ mouse model has shown that very high levels of Ag-specific CD8⁺ T cells are generated and maintained into memory (9). The size of the CD8⁺ T cell memory pool is reflective of the initial clonal burst size during the acute LCMV response (10). The TCR repertoire of the primary antiviral CD8 T cell response to LCMV was shown to be similar to that of the memory pool (11). Altogether, these data suggest a stochastic selection of memory cells from the pool of CD8⁺ T cells activated during primary infection.

Similar conclusions were reached by Callan et al. (12) in a study of CD8⁺ T cell responses from acute through chronic EBV infection. However, recent evidence from human HIV-1 infection has shown that an HLA-A2-restricted response to an HIV-1 gag epitope found in a majority of HLA-A2 individuals with chronic HIV-1 infection was not present in 11 HLA-A2 individuals during acute HIV-1 infection (13). This suggests that for persistent viral infections the memory/effector pool may not be reflective of the initial pool of CD8⁺ T cells and that the generation of CD8⁺ T cells may not fit a stochastic model.

To examine the generation and maintenance of CD8⁺ T cell responses during a persistent human viral infection, we analyzed a cohort of individuals experiencing acute EBV infection using several EBV peptide epitopes restricted by common HLA alleles: HLA-A*0201 (A2); HLA-B*0702 (B7); and HLA-A*0301 (A3). By studying individuals with acute EBV infection, we were able to determine the frequencies of CD8⁺ T cell responses to previously defined lytic and latent epitopes during acute infection. We then followed these same individuals into latent EBV infection to characterize the evolution of EBV epitope-specific CD8⁺ T cell responses over time.

These studies were conducted in adolescents (17–24 yr old) presenting to the clinic at the University of Massachusetts Amherst Student Health Service (Amherst, MA) with clinical symptoms consistent with acute EBV infection (fever, rash, fatigue, hepatosplenomegaly). After informed consent, students presenting with acute-like symptoms gave five blood samples (50 ml each) at the following time points: at presentation with symptoms (V-1); and 1 wk (V-2), 2 wk (V-3), 6 mo (V-4); and 1 yr (V-5) after presentation. Inclusion in the study was based on a positive Monospot test and the presence of atypical lymphocytes. Acute EBV infection was confirmed through the detection of IgM for the EBV viral capsid Ag.

Healthy EBV-seropositive adults >30 yr of age were recruited for these studies from the UMMS research community. All of these individuals likely experienced EBV infection for a minimum of 10 yr before study. Prior EBV infection was confirmed through the detection of EBV capsid Ag-specific IgG Abs in the peripheral blood. After informed consent, study subjects provided blood samples (50 ml) at entry and every 3 mo thereafter. These studies were approved by the Human Studies Committee at the University of Massachusetts Medical School (Worcester, MA).

Molecular HLA class I typing was done on all study participants by Microdiagnostic (Nashville, TN).

A modification of the method by Kern et al. (14) was used for the detection of cytokine-secreting CD8⁺ T cells. Whole blood (0.3 ml/test, heparin) was incubated for 1 h at 37°C with 2 μM EBV peptides. Because staphylococcal enterotoxin B (Toxin Technology, Sarasota, FL) is able to nonspecifically stimulate T cells to secrete IFN-γ, it was used as a positive control. None of the donors used in this study was HIV-1 infected; therefore, the HLA-A2-binding, HIV-1 gag SLYNTVATL peptide was used as a negative control. After 1 h incubation with peptide, Golgiplug (BD PharMingen, San Diego, CA) was added to the cells, and they were incubated for an additional 5 h. After incubation, the cells were incubated with 2 μM EDTA for 15 min with vigorous vortexing every few minutes. The cells were stained with combinations of the following Abs: IFN-γ APC, CD3 PerCP, CD69 PE, and CD8 FITC. The cells were incubated at room temperature for 30 min and then washed with 1% FBS in PBS; 100,000 events gated on the lymphocyte population were collected and analyzed immediately by 4-color flow cytometry. Appropriate isotype, negative, and positive controls were used to define positive and negative cell populations. Background IFN-γ production in this assay was 0.02 ± 0.01% for stimulation of individuals with either a peptide derived from HIV-1- or EBV-derived non-HLA-binding peptides. Significant IFN-γ production was considered to be greater than the mean of the background plus 3 SD, or 0.05% of CD8⁺ T cells.

HLA-A*0201 H chain (residues 1–275) and human β₂-microglobulin (β₂m) in the prokaryotic expression system pET R&D were obtained from D. Garboczi (National Institutes of Health, Bethesda, MD). The 3′ end of the HLA-A*0201 H chain was modified with a BIR A biotinylation site as published (15). HLA-B*0702 was a gift of Dr. E. Pamer (Memorial Sloan-Kettering Cancer Center, New York, NY). As previously described (16), HLA class I H chain constructs or β₂m were grown to midlog phase and induced with 0.5 M isopropyl-β-thiogalactosidase. Inclusion bodies were purified and solubilized in 6 M guanidine-HCl, pH 8.2.

The H chain, β₂m, and peptide were refolded by dilution as described (16). The A2 BMLF-1 peptide, GLCTLVAML, and the B7 EBNA-3A peptide, RPPIFIRRL, were synthesized by the University of Massachusetts Medical Center peptide core facility and purified before usage. Briefly, 1 μM H chain, 2 μM β₂m, and 30 μM peptide were refolded by dilution (0.4 M l-arginine, 0.1 M Tris (pH 8.0), 0.002 M EDTA, 0.005 M reduced glutathione, 0.5 M oxidized glutathione, and 0.5 mM PMSF (pH 8.3)) and incubated for 24 h at 10°C with gentle stirring. The refolded monomer was concentrated using an Amicon Stirred Cell (Amicon, Beverly, MA). The 45-kDa refolded product was isolated on an ACTA fast protein liquid chromatograph using a Superdex 200 column (Amersham-Pharmacia, Piscataway, NJ). After gel filtration, the refolded monomer was biotinylated with biotin ligase (birA; Avidity, Denver, CO) and washed to remove excess biotin. The resulting biotinylated monomer was mixed with streptavidin-APC conjugate (BD PharMingen) at a molar ratio of 4:1 to form tetramers (15). To stain 100 μl whole blood, 100 ng tetramer were used. Percentages of tetramer-binding cells were expressed as a percentage of total CD8⁺ T cells. These tetramers stained non-A2 or non-B7 individuals with acute EBV infection and A2, B7 individuals who were EBV seronegative at 0.02 ± 0.01%. Tetramer staining greater than the mean of the background plus 3 SD, or 0.05% of CD8⁺ T cells, was considered significant. A total of 100,000 events gated on the lymphocyte population were collected and analyzed.

Four-color flow cytometric analysis was performed using whole EDTA blood for Ki-67 analysis and heparin preserved blood for annexin V analysis. Because annexin V binding is Ca²⁺ dependent, EDTA blood stained with annexin V was run alongside as a negative control. Samples were analyzed on a BD Biosciences FACSort with an added laser and FACSCaliber software (BD Biosciences, San Jose, CA). HLA class I tetramers can bind nonspecifically to non-T cell populations; therefore cells were first gated through both CD3 and CD8 to ensure that the tetramer positive cells were all CD8⁺ T cells. Cells were analyzed with a tetramer conjugated to APC, CD8 FITC (Sigma, St. Louis, MO), CD3 PerCP (BD Biosciences), and Ki-67 PE or annexin V PE (BD PharMingen, San Diego, CA). Permeabilization for staining with Ki-67 was done after tetramer staining and lysing of the cells. Permeabilized isotype controls were run alongside each analysis for Ki-67.

Means and SDs were calculated and compared with each other using the Student t test to determine p values. The Wilcoxon signed rank test, a nonparametric paired t test, was used to determine p values for the differences in the range of tetramer-staining CD8⁺ T cells between study time points for the group with acute EBV infection. The Mann-Whitney U test was used to determine p values between the range of tetramer staining cells between study time points for the acute EBV infection group and the long term latently infected group. To determine whether differences in the percentage of tetramer staining cells were significant, tetramer staining was performed on 12 identical samples for 14 individuals, and the means ± SD for those 12 samples were determined. These were then compiled for the 14 individuals to determine normal variations in tetramer staining. Differences >0.2 ± 0.2% were considered to represent a significant change in magnitude.

We first examined the frequency and magnitude of CD8⁺ T cell responses to previously defined EBV lytic protein epitopes (Table I) during acute and latent EBV infection using an in vitro assay that detects single-cell production of IFN-γ (Fig. 1). The three most prevalent HLA types in our acute population were HLA-A2⁺ (45%, n = 25), HLA-A3⁺ (16%, n = 10), and HLA-B7⁺ (23%, n = 12); 67% of our acute population had one or more of these alleles. We therefore prioritized evaluation of responses to these epitopes. Because HLA typing results were not available at baseline testing, we used all of the peptides listed in Table I during presentation with symptoms (V-1); once HLA typing was completed, peptide panels were individualized for later visits. We then followed these individuals with acute EBV infection into latent EBV infection and evaluated their CD8⁺ T cell responses up to 1 yr postpresentation (V-5). Fresh whole blood was used for all assays to avoid the possible loss of Ag-specific CD8⁺ T cells during Ficoll (Amersham Pharmacia Biotech, Piscataway, NJ) separation and cell cryopreservation.

The frequencies and epitope specificity of peripheral blood CD8⁺ T cell responses were evaluated from presentation through 1 yr and were compared with those detected in long term latently infected individuals. Of the eight early or lytic protein epitopes evaluated, all but one were recognized by the majority of individuals tested (Fig. 2). CD8⁺ T cell responses to an HLA-A2⁺-restricted BMLF-1 epitope were detected in the majority of HLA-A2⁺ individuals during acute and latent infection. An HLA-A2⁺-restricted gp85 epitope (gp85SV) was also recognized during the acute and the latent stages of infection. By contrast, CD8⁺ T cell responses to the six other lytic epitopes were more commonly detected during acute infection than during latent infection.

All HLA-A2⁺ individuals with acute EBV infection produced IFN-γ in response to stimulation with the A2 BMLF-1 peptide (range, 0.08–12%; median, 2.3%; Fig. 2). The majority (25 of 27 = 93%) of latently infected individuals also produced IFN-γ in response to this peptide at frequencies ranging from 0.06 to 1.6% (median, 0.56%). Responses to the A2 BZLF-1 epitope were commonly seen during acute EBV infection (5 of 8 = 63% of individuals; range of IFN-γ production, 0–1.25%) but rarely detected during latent infection (1 of 8 = 12.5%; range of IFN-γ production, 0–0.15%). Eight of ten (80%) HLA-B7⁺ individuals responded to the B7 BZLF-1 peptide (range, 0–1.4%) during acute infection; the response to the B7 BZLF-1 peptide was detected in only 25% of individuals (2 of 8; range, 0–0.34%) during latent infection. Similar to the pattern seen for HLA-A2⁺- and HLA-B7⁺-restricted lytic epitopes, responses were also more frequently detected to the A3 BRLF-1 epitope during acute EBV infection (8 of 10 = 80%; range, 0–3.2%), than during latent EBV infection (1 of 8 = 13%; range, 0–0.03%).

We also examined the frequency and magnitude of CD8⁺ T cell responses to previously defined EBV latent protein epitopes (Table I) in acute and latent EBV infection using the in vitro assay for the production of IFN-γ (Fig. 3). Latent epitope responses were rarely detected in individuals presenting with acute EBV infection. Only 2 of 20 (10%) HLA-A2⁺ individuals responded to the A2 EBNA-3A epitope during acute EBV infection (range, 0.05–0.06%), but 14 of 20 (70%) latently infected HLA-A2⁺ individuals responded to the A2 EBNA-3A epitope (range, 0.05–0.41%); 3 of 20 (13.6%) latently infected HLA-A2⁺ individuals also responded to the A2 LMP-1 epitope (range, 0.09–0.5%). Whereas only 3 of 12 (25%) acutely infected HLA-B7⁺ individuals responded to the B7 EBNA-3A epitope (range, 0–0.44%), all 10 latently infected individuals responded to the B7 EBNA-3A epitope (range, 0.08–1.2%). A similar pattern was detected for the B7 EBNA-3C epitope. Only one HLA-A3⁺ individual (n = 8; 0.23%) responded to the A3 EBNA-3A epitope during acute EBV infection, and this response was not maintained into latent EBV infection.

We detected IFN-γ responses to the A2 BMLF-1 peptide in all individuals with acute EBV infection and in all of our acute study individuals 1 yr post-EBV infection. Because IFN-γ production can underestimate the frequency of these responses, we developed HLA class I tetramers to more precisely determine the magnitude of the A2 BMLF-1-specific response (Fig. 4,a). At presentation with acute EBV infection, A2 BMLF-1 (lytic EBV gene) tetramer-binding CD8⁺ T cells were detected in all HLA-A2⁺ individuals (V-1, n = 25; range, 0.1–47% of CD8⁺ T cells; median, 2.31% (Fig. 4 b). One week postpresentation (n = 19), the percentage of A2 BMLF-1-specific T cells ranged from 0.2 to 15.9% (median, 2.35%); 7 of 19 (36%) HLA-A2⁺ individuals at this time point had increases in their percentages of A2 BMLF-1 tetramer-binding CD8⁺ T cells. The median frequency of A2 BMLF-1 tetramer-binding CD8⁺ T cells then decreased at 2 wk (V-3; n = 19; range, 0.14–9.76%; median, 1.08%), 6 mo (V-4; n = 11; range, 0.2–1.93%; median, 1.19%) and 1 yr postpresentation (V-5; n = 7; range, 0.08–1.4%; median, 0.23%).

A2 BMLF-1 tetramer-binding CD8⁺ T cells were also detected in 20 of 22 (91%) HLA-A2⁺ healthy long term EBV seropositive individuals studied (range, 0.05 to 1.95%; median, 0.32%). The frequencies of A2 BMLF-1 tetramer-binding CD8⁺ T cells in the acute EBV-infected individuals 1 yr postpresentation were significantly different than the frequencies of A2 BMLF-1 tetramer-binding CD8⁺ T cells detected in acute patients at presentation (p = 0.02), 1 wk (p = 0.02), and 2 wk (p = 0.05) but not different than the frequencies of A2 BMLF-1 tetramer-binding CD8⁺ T cells detected at 6 mo. The frequencies of A2 BMLF-1 tetramer-binding CD8⁺ T cells in the long term latent group were significantly different from the range of A2 BMLF-1 tetramer-binding CD8⁺ T cells in the acute patients at presentation (p = 0.001), 1 wk (p = 0.001), and 2 wk (p = 0.01) but not significantly different from the frequencies of tetramer-binding CD8⁺ T cells at 6 mo or 1 yr.

B7 EBNA-3A-specific responses were detected in only three of eight of the HLA-B7⁺ individuals with acute EBV infection but in all EBV latently infected individuals. We developed a B7 EBNA-3A tetramer to more precisely determine the magnitude of epitope-specific CD8⁺ T cells during acute and latent EBV infection (Fig. 5,a). B7 EBNA-3A tetramer-binding CD8⁺ T cells were detected rarely (3 of 9 = 33%) and at low frequencies in HLA-B7⁺-positive individuals with acute EBV infection at multiple time points within 2 wk of symptom onset (V-1, V-2, or V-3; range, 0–0.44%; Fig. 5 b). In the three individuals with detectable B7 EBNA-3A tetramer-binding CD8⁺ T cells during V-1, the changes in magnitude between presentation and 2 wk were not statistically different. B7 EBNA-3A tetramer-binding cells have been detected in all individuals 6 mo after entry into the study (V4; 8 of 8; range, 0.08–1.2%; median, 0.18%) and 1 yr after entry to the study (V5; 5 of 5; range, 0.16–0.46%; median, 0.24%). B7 EBNA-3A tetramer-staining cells have also been detected in all 6 HLA-B7⁺ EBV-long term-seropositive individuals studied thus far (range, 0.08–1.95%; median, 0.39%).

During the first 2 wk after presentation with acute EBV infection, a decrease in the frequency of A2 BMLF-1-specific CD8⁺ T cells was noted (Fig. 4,b). By contrast, there was very little change in the percentages of B7 EBNA-3A CD8⁺ T cells for the three individuals analyzed (Fig. 5,b). We wanted to determine whether the initial burst size of CD8⁺ T cells determined the final magnitude of response detected 1 yr post-EBV infection. Comparison of the percentages A2 BMLF-1-specific CD8⁺ T cells in early acute infection with the percentage of A2 BMLF-1-specific CD8⁺ T cells at 1 yr for the same individual revealed no correlation between the initial and final magnitude of the response (Fig. 4,b and data not shown). Additionally, there appeared to be no detectable burst for the B7 EBNA-3A tetramer-binding CD8⁺ T cells; there were no significant differences between the initial and final magnitude of the B7 EBNA-3A response in the 3 HLA-B7⁺ individuals studied (Fig. 5 b).

Because the early A2 BMLF-1-specific CD8⁺ T cell frequencies did not appear to predict CD8⁺ T cell frequencies at 6 mo or 1 yr, we wanted to determine the dynamics of loss or gain of CD8⁺ T cells during the year. Tables II and III show the fold loss (or increase represented by a number <1) over time in the frequencies of A2 BMLF-1 or B7 EBNA3A tetramer-staining cells. Large changes in the frequencies of A2 BMLF-1 tetramer-binding CD8⁺ T cells were observed; in most (6 of 10) individuals, the largest changes occurred between 2 wk and 6 mo postpresentation. By contrast, the frequencies of the B7 EBNA-3A tetramer-binding CD8⁺ T cells did not change much over the course of our study. All of the HLA-B7⁺ individuals who did not have detectable B7 EBNA-3A tetramer-binding CD8⁺ T cells during the first 2 wk of our study developed them between 2 wk and 6 mo of the study.

Further analysis of the dynamics of loss or gain of A2 BMLF-1 and B7 EBNA-3A tetramer binding CD8⁺ T cells was done directly ex vivo on whole EDTA blood by four-color flow cytometry using Ki-67, a marker for cellular proliferation, or annexin V, a marker for cell death. Fig. 6 shows analyses of the expansion of epitope-specific cells using Ki-67, an Ag expressed during all active stages of the cell cycle. At presentation with symptoms, 60 ± 12% of the A2 BMLF-1 tetramer-binding CD8⁺ T cells coexpressed Ki-67. A significant reduction in Ki-67 coexpression was noted at 1 (26 ± 16) and 2 wk postpresentation (5 ± 4%) and persisted through 6 mo to 1 yr postinfection (Fig. 6 a). Analyses of Ki-67 expression for 3 HLA-B7⁺ individuals with detectable B7 EBNA-3A-specific CD8⁺ T cells at V-1–V-3 showed similar Ki-67 expression (data not shown). The coexpression of Ki-67 on the A2 BMLF-1 tetramer-binding CD8⁺ T cells was not significantly different from the coexpression of Ki-67 in the whole CD8⁺ T cell population at all time points.

Annexin V staining can be used to detect the translocation of phospholipid phosphatidylserine from the inner to the outer cell membrane, an early stage in apoptosis. Analysis of A2 BMLF-1 tetramer-binding CD8⁺ T cells showed that at presentation, 7.43 ± 7.95% coexpressed annexin V (Fig. 6 b). The percentage of CD8⁺ T cells coexpressing annexin V appeared to decrease at 1 wk (3.59 ± 4.42%) and 2 wk (2.04 ± 2.47%) after presentation, but these differences were not statistically significant. Similar patterns of expression of annexin V were detected in B7 EBNA-3A tetramer-binding CD8⁺ T cells. No annexin V expression was detected in the A2 BMLF-1 or B7 EBNA-3A tetramer-binding CD8⁺ T cells from the 6-mo or 1-yr postacute EBV infection patients or for the long term-seropositive individuals.

EBV is a chronic viral infection in which Ag persists during the lifetime of the host, and the effect of this Ag on the immune system over time is unknown. Selin et al. (27) have shown that the frequency of virus-specific CD8⁺ T cells can change after infection with heterologous viruses. Humans do not live in germfree environments, and we would expect our long term EBV-seropositive donors to come into contact with many different types of pathogens during a period of 2 yr. To determine the stability of the EBV-specific CD8⁺ T cell populations over time, we monitored eight long term EBV-seropositive individuals several times during a period of 20 mo to determine their change in frequency of A2 BMLF-1- or B7 EBNA-3A-specific CD8⁺ T cell responses. Fig. 7 demonstrates that the frequency of both A2 BMLF-1- and B7 EBNA-3A-specific CD8⁺ T cell responses were stable over time.

This study is among the first to study the evolution of epitope-specific CD8⁺ T cell responses in a large cohort of individuals with acute through latent EBV infection. We have shown that several epitope-specific immune responses generated during acute EBV infection were not maintained into latent EBV infection while other epitope-specific CD8⁺ T cell responses that were not dominant during acute EBV infection were generated and/or maintained during latent EBV infection. Additionally, using HLA class I tetramers we have shown that early epitope-specific CD8⁺ T cell frequencies did not correlate with the magnitude of the response at 6 mo or 1 yr after EBV infection. Altogether, these data suggest that the CD8⁺ T cell repertoire generated during a persistent viral infection is not simply reflective of the initial pool of CD8⁺ T cells generated during acute infection. Interestingly, the percentages of A2 BMLF-1- and B7 EBNA-3A-specific CD8⁺ T cells in the eight long term EBV-seropositive individuals studied during a 2-yr period were constant, suggesting that an equilibrium between virus and the immune response had been reached.

Of the eight early or lytic protein epitopes evaluated, all but one (an HLA A2-restricted gp350 epitope) were recognized by the majority of individuals evaluated during acute EBV infection. By contrast, responses to eight previously defined EBV latent protein epitopes were infrequently detected during acute infection. Of the eight EBV latent protein epitopes evaluated, only one A2- and two B7-restricted EBNA-3 epitopes were well recognized during latency. When detected, the CD8⁺ T cell responses to EBV latent protein epitopes were of lower frequency (≤1% of CD8⁺ T cells) than CD8⁺ T cell responses to the lytic epitopes and remained stable during the 1 yr of follow-up. These results suggest the predominance of lytic epitope-specific CD8⁺ T cell responses during acute EBV infection and are in agreement with other recently published data from fewer individuals (23, 28, 29, 30). Although it has been suggested that the differential response of CD8⁺ T cells to lytic and latent Ags is due to different stages of the EBV lifecycle (13), all of the latent gene products are expressed during the lytic cycle (31). Lytic proteins also appear to be expressed periodically during latent EBV infection (32, 33). Recently, it has been reported that although the EBNA-3A protein is expressed by cultured BLCL lines in vitro, EBNA-3A expression was not detected in either the peripheral blood or tonsillar cells of latently infected healthy, EBV-seropositive donors (34). Although these data do not exclude intermittent or ongoing low level expression of EBNA-3A (particularly at a site outside the blood), they suggest that the differential expression of viral proteins does not fully account for the response to these Ags.

Of interest is that Callan et al. (12, 28, 29) have reported that HLA-B8⁺ individuals generate and maintain an HLA-B8⁺-restricted CD8⁺ T cell response to a BRLF-1 epitope. In all but one of our HLA-A3⁺ individuals, BRLF-1 epitope-specific immune responses were readily detectable during acute EBV infection, but not at 1 yr post-EBV infection. If the presence of a specific CD8⁺ T cell response were simply dependent on the presence of Ag, we would expect to see maintenance of BRLF-1 responses to both the A3 and the B8 epitopes. Our data demonstrating loss of an HLA-A3⁺-restricted CD8⁺ T cell response to BRLF-1 along with the data of Callan et al. demonstrating the maintenance of an HLA-B8-restricted response to BRLF-1 suggest that maintenance of the response to this epitope may be determined in a large part by the HLA type of the individual. Alternatively, cross-reactive responses present before antigenic challenge may lead to preferential expansion and maintenance of some responses compared with others. This has been demonstrated in the mouse model of acute LCMV infection, in which pre-existing responses to LCMV resulted in increased CD8⁺ T cell frequencies to heterologous viral epitopes when compared with naive mice (35).

Study of the dynamics of the generation and maintenance of BMLF-1 (lytic) and EBNA-3A (latent) epitope-specific CD8⁺ T cells using HLA class I tetramers showed that the early frequencies of CD8⁺ T cell responses to these epitopes did not correlate with their maintenance into the pool of memory CD8⁺ T cells. Previous studies in the mouse have shown a correlation between the initial burst size and the final magnitude of the response (10). Our results demonstrated a massive expansion of A2 BMLF-1-specific CD8⁺ T cells, but large expansion of B7 EBNA-3A-specific CD8⁺ T cells, or any of the other latent EBV-specific CD8⁺ T cells were not detected during any of the time points of this study. These EBNA-3A-specific CD8⁺ T cell responses were maintained at percentages similar to those of the A2 BMLF-1-specific CD8⁺ T cells in all of the HLA-A2⁺B7⁺ individuals. These data suggest that a large burst size is not necessary to provide the host with both effector and long term memory CD8⁺T cells and that the generation of memory CD8⁺ T cells in a persistent viral infection is not simply a stochastic process.

In contrast to the dynamic nature of epitope-specific CD8⁺ T cell responses during acute infection, the frequency of A2 BMLF-1- and B7 EBNA-3A-specific CD8⁺ T cells remained constant in all chronically infected individuals studied during 2 yr. Although these individuals likely experienced many immune challenges during the 2 yr of study, there were only minor perturbations in their frequencies over time. Data generated by Khan et al. (36) demonstrated that EBV viral DNA loads are stable in individuals over time periods spanning several years. Periodic reactivations of the EBV virus in the peripheral blood have also been demonstrated, suggesting that the stability DNA load over time is not that way because the virus is quiescent (32, 33). Our data demonstrating that CD8⁺ T cell frequencies to two different EBV epitopes are also stable over time suggest that during persistent viral infections a long term homeostasis exists between the virus and the immune system.

In summary, the study of EBV as a model for the generation and maintenance of epitope-specific CD8⁺ T cells in a human persistent viral infection has shown that the memory pool is not reflective of the initial pool of CD8⁺ T cells and that the generation of CD8⁺ T cell responses may not fit a stochastic model. The maintenance of epitope-specific CD8⁺ T cells to a persistent viral infection over time reflects a homeostasis reached between virus and the immune system. Perturbations in this homeostasis may in part explain EBV-associated diseases.

We thank Rose Cicarelli for her assistance with obtaining patient samples at the UMass-Amherst acute EBV clinic. We are also indebted to Drs. David Garboczi (National Institutes of Health), Marcelo Kuroda, and Joern Schmitz (Harvard Medical School) for helping to set up tetramer technology in our laboratory and to Dr. Eric Pamer (Memorial Sloan-Kettering Cancer Center) for the B*0702 HLA class I H chain construct. We thank Drs. Liisa Selin and Raymond Welsh for helpful discussions and review of the manuscript.