Changing Patterns of Dominant TCR Usage with Maturation of an EBV-Specific Cytotoxic T Cell Response¹

The CD8⁺ CTLs play an important role in the control of virus infections, recognizing infected cells through the presence on the target cell surface of viral peptides complexed with MHC class I molecules (1). Specific recognition of such complexes is mediated via the TCR, a membrane-bound heterodimer composed of unique α- and β-chains with variable (V), diverse (D; seen in β-chains only), joining (J), and constant (C) regions. Several factors contribute to TCR repertoire diversity, namely, the multiple possible V(D)J combinations generated during TCR gene rearrangement, the random mutations or nucleotide additions introduced at V(D)J junctions, and the random pairings of separately rearranged α- and β-chains (2). Individual V gene choice and in particular the structure of the complementarity-determining region (CDR)³ 3 encoded by V(D)J junction sequences are thought to be critical determinants of TCR specificity (2, 3, 4).

Primary virus infections, both in mouse (5, 6, 7) and the human (8, 9, 10), are often characterized by large expansions of activated CD8⁺ cytotoxic T cells. Within these one can find preferentially expanded Vβ subsets, often with oligoclonal TCR usage (11, 12, 13, 14), implying that such proliferations are Ag driven rather than the result of bystander T cell activation. However, it is only recently that single-cell assays of MHC/peptide tetramer staining and of peptide-induced cytokine release have shown the extent to which a few viral epitope-specific reactivities can dominate the acute CD8⁺ population (15, 16, 17, 18). The sheer size of this primary response contrasts with the situation subsequently found in memory, where virus-specific reactivities are again detectable both by single-cell assays ex vivo and by functional analysis of in vitro reactivated T cell preparations but now at much lower frequencies.

The lineage relationship between primary and memory responses, and the factors which determine selection into the memory repertoire, remains poorly understood (19, 20). Experiments in murine model systems have mainly focused on agents such as lymphocytic choriomeningitis virus (LCMV) (11, 12, 15, 17, 21) or Listeria monocytogenes (22), which are rapidly cleared by the CTL response. In this study, the composition of specific T cell memory, in the absence of further antigenic challenge, appeared to reflect that seen in the acute primary response. However, the response to rechallenge ranged in different studies from no detectable change between primary and secondary effector populations (12) to a narrowing of the original pathogen-specific repertoire (22), or to varying degrees of repertoire diversification and the emergence of new secondary clonotypes (11, 21). This latter pattern was accentuated under the unusual conditions of high-dose primary infection with the LCMV mutant c13, where failure of the immune response to control the infection was associated with rapid sequential exhaustion of virus-specific clonotypes (11). Clonotype loss has also been reported in patients undergoing primary HIV infection (23), another situation where there is overwhelming virus replication throughout the lymphoid system. Here, however, interpreting prospective change in the virus-specific CTL repertoire is complicated not only because new viral variants are continually emerging in vivo (24) but also because HIV itself targets the CD4⁺ T cell pool, upon which the successful maintenance of CD8⁺ memory is thought to depend (25).

The present work examines maturation of the human CD8⁺ T cell response to EBV, a genetically stable B lymphotropic agent which induces a clinically obvious primary infection with marked CD8⁺ T cell activation (infectious mononucleosis (IM)) but subsequently persists for life as a low-grade asymptomatic infection in balance with host immune surveillance (26). Most work in the EBV system to date has concerned CD8⁺ T cell responses to virus-latent cycle Ags, in particular to the EBV-encoded nuclear Ags (EBNAs) 3A, 3B, and 3C which are preferentially recognized by the memory T cell response in virus carriers (27, 28, 29). Interestingly, one of the best-studied memory responses, to the HLA-B8-restricted FLRGRAYGL epitope (designated FLR) in EBNA 3A, is characterized by a highly focused use of particular TCR clonotypes (30). This prompted studies on HLA-B8-positive IM patients showing that the same dominant clonotypes were also abundant in the primary CTL response and suggesting that the composition of the FLR-specific repertoire does not vary dramatically over time (31, 32).

Recent work has shown, however, that individual latent Ag-specific reactivities account for only a very small fraction (rarely above 1%) of the total CD8⁺ response in IM patients (18, 33). By contrast, reactivities to immunodominant lytic Ag-derived epitopes (34) such as the B8⁺-restricted RAKFKQLL epitope (designated RAK) from the immediate early lytic cycle protein BZLF1 or the A2.01-restricted GLCTLVAML epitope (designated GLC) from the early lytic cycle protein BMLF1 are at least 10-fold more abundant as visualized by tetramer staining (18). It remains to be seen 1) whether these much larger responses to lytic cycle Ags can also be dominated by just a few clonotypes and in that way explain some of the preferential Vβ subset expansions seen among the CD8⁺ T cell pool in IM (14), and 2) if that were the case, whether highly expanded T cell clonotypes present within the primary response remain dominant in the longer term. An earlier study of in vitro-derived clones from the RAK-specific response in IM and post-IM patients indicated the involvement of a diverse range of TCRs during primary infection and persistence (35). Indeed, the overall strength of this response has been attributed to its multiclonal composition, a trend that has also been noted with some (but not all) immunodominant EBV-latent cycle epitopes (36). The present work describes a detailed clonotypic analysis of primary and memory responses to the immunodominant lytic cycle epitope GLCTLVAML (BMLF1 residues 280–288).

Patients with acute IM, identified on clinical grounds and by heterophile Ab positivity, were sampled within 7–15 days of the onset of clinical symptoms. Heparinized blood (30–60 ml/patient) was processed by standard methods (33) to prepare PBMC; one aliquot of cells was used for HLA class I genotyping, another aliquot was used to generate a lymphoblastoid cell line (LCL) by B95.8 strain EBV-induced transformation in vitro, and the remaining aliquots were placed in cryostorage until required. All patients showed a complete recovery from disease symptoms within 3–4 wk; the same individuals were rebled on a second occasion 12–40 mo later (mean interval, 26 mo) and PBMC were again prepared and cryopreserved before analysis. A healthy HLA-A2.01-positive individual CMc, who had been EBV seropositive for at least 10 years and who had no history of IM, was included as a reference donor in the analysis of GLC-specific memory.

The PE-conjugated HLA-A2.01/GLC peptide tetrameric complex was prepared as described previously (18). As in earlier work, the specificity of A2.01/GLC tetramer staining was confirmed by testing first on reference GLC-specific T cell clones vs control clones specific for other A2.01-restricted EBV epitopes and second on reference PBMC from A2.01-positive EBV-seropositive donors vs control PBMC from A2.01-positive EBV-seronegative donors and A2.01-negative IM patients (18). As a further control, the IM PBMC populations tested here were shown not to stain with irrelevant tetramers, e.g., an A2.01/HIV gag epitope tetramer and other EBV epitope tetramers, including B8/RAK and B8/FLR tetramers (18), involving HLA class I alleles other than A2.01. For two-color staining, 10⁶ PBMC were incubated on ice for 30 min in PBS + 0.1% BSA + 0.1% sodium azide containing 0.5 μg/ml PE-conjugated tetramer and saturating amounts of a Tricolor-conjugated anti-CD8 mAb (mAb, isotype IgG2a; Caltag, South San Francisco, CA). For three-color staining, the above samples were additionally incubated on ice for another 30 min with one of a panel of anti-human Vβ-chain mAbs (Immunotech, Marseille, France): E2.2E7.2 (Vβ2-specific, IgG1), TAMAYA1.2 (Vβ16-specific, IgG1), BA62.6 (Vβ18-specific, IgG2b), and IMMU546 (Vβ22-specific, IgG1). Of these, the Vβ16- and Vβ22-specific mAbs were directly FITC conjugated, whereas binding of the unconjugated Vβ2- and Vβ18-specific mAbs was detected using a FITC-labeled goat anti-mouse Ig isotype Ab (Southern Biotechnology Associates, Birmingham, AL). Parallel studies with the same set of Vβ subset-specific mAbs, in combination with the anti-CD8 mAb, have been conducted on PBMC from a panel of 20 healthy control donors to establish mean values (±SD) for percent Vβ representation within the circulating CD8⁺ T cell pool; the relevant values were 5.01 ± 2.09% for Vβ2, 0.69 ± 0.3% for Vβ16, 0.27 ± 0.43% for Vβ18, and 2.62 ± 1.02% for Vβ22 (J. Faint, unpublished observations). Some studies also included staining with a recently developed Vβ4-specific mAb (IgM, kindly provided in FITC-conjugated form by F. Romagne, Immunotech). Note that before the analysis of double- or triple-stained PBMC samples, the flow cytometer was carefully color-compensated using aliquots of the same PBMC singly labeled with PE-conjugated or FITC-conjugated anti-CD4 mAbs (Becton Dickinson, San Jose, CA) or with the Tricolor-conjugated anti-CD8 mAb. Then the quadrant boundaries were set using aliquots of the same PBMC stained with the relevant isotype-matched control mAbs (Dako, Cambridge, U.K.) either directly or indirectly labeled as appropriate.

Primary CTL clones were established either by limiting dilution of thawed IM PBMC directly into IL-2-enriched medium at 0.3 and 3 cells/0.2-ml round-bottom well as previously described (33) or by FACS of GLC tetramer-positive cells directly into 96-well plates at 1 cell/well in the presence of irradiated preactivated allogeneic PBMC feeders (10⁶/ml) with irradiated autologous LCL cells (10⁵/ml). Growing microcultures were further expanded by transfer into 2-ml wells using the same stimulation protocol as before. Feeder cells were from pooled fresh buffy coats (National Blood Service, Birmingham, U.K.) and were incubated with PHA at 10 μg/ml for 1 h and then washed five times before irradiation and use. Memory CTL clones were established using the optimal protocol for peptide-induced reactivation of human T cell memory originally developed for the EBV system by Lalvani et al. (37) and subsequently confirmed as the most efficient means of accessing the GLC-specific response (N. E. Annels, unpublished observations); note that for CTL memory to an EBV-lytic cycle epitope such as GLC, the epitope peptide is a much more effective in vitro stimulus than are latently infected LCL cells. Thus, PBMC from post-IM donors and from the reference donor CMc were exposed to the synthetic peptide GLCTLVAML at the optimal concentration of 100 μg/ml (37) and, after a 2-h incubation at 37°C in 5% CO₂, seeded at 2 × 10⁶/ml in IL-7-conditioned medium. Recombinant IL-2 was added at 10 U/ml on day 3, after which limiting dilution cloning was conducted as above. Procedures for peptide synthesis and solution in DMSO were as previously described.

CTL clones specific for the GLC epitope were identified by the functional analysis in standard chromium release cytotoxicity assays. The target cells were autologous or HLA-A2-matched LCL pretreated with 0.2 μg/ml GLC peptide or, as a control, with an equivalent concentration of DMSO solvent, and preinfected with a recombinant vaccinia virus vacc-BMLF1 expressing the BMLF1-lytic cycle Ag from with the epitope is derived or, as a control, with the vacc-TK-recombinant (34).

mRNA was extracted from T cell clones with confirmed GLC specificity using Tri Reagent (Boehringer Mannheim, Indianapolis, IN). cDNA was synthesized, tailed with poly(G), and used as a template in anchored PCR containing a 3′ Cα primer containing a SalI site (TGACCGCAGTCGACAGACTTGTCACTGGATT) or a 3′ Cβ primer containing a SalI site (ATACTGGAGTCGACGGAGATCTCTGCTTCTGATG) and a 5′ poly(C) primer containing a NotI site (GCATTCAGCTGCGGCCGCCCCCCCCCCCCCCCC). The PCR conditions were 94°C for 4 min, followed by 5 cycles of 94°C for 1 min, 65°C for 1 min, and 72°C for 2 min, followed by 30 cycles of 94°C for 1 min, 58°C for 1 min, and 72°C for 2 min. The DNA product was cloned into a Phagescript M13 vector (Stratagene, La Jolla, CA). After transformation of Escherichia coli (strain XL1 blue MRF′), plaques containing inserts were sequenced. Only those clones found to express a single α- and/or β-chain were included in the study. The nomenclature used for the TCR elements is as described by Arden et al. (38).

The clonotypic composition of the primary GLC epitope-specific response was studied by in vitro cloning from three HLA-A2.01-positive IM patients, IM74, IM61, and IM69. Each of these had a marked elevation of CD8⁺ T cell counts in the blood and had shown functional evidence of GLC-specific cytotoxicity when PBMC taken at the acute stage of the disease were tested immediately ex vivo on GLC peptide-loaded targets (34). The same three individuals were also sampled on a second occasion over 2 years later, long after the resolution of the disease, and the GLC-specific memory T cell repertoire again studied by cloning.

IM74 was studied in the acute phase of primary infection, at a time when 5.6% circulating CD8⁺ T cells stained with the A2.01/GLC tetramer; in this case, GLC-specific T cell clones were established from the acute phase blood sample both by limiting dilution culture of the PBMC population and by single-cell cloning of cells FACS sorted on the basis of their staining with the A2.01/GLC tetramer. All clones were expanded by culturing on the autologous LCL and pooled allogeneic feeders in IL-2-conditioned medium, and then tested for Ag- and epitope-specific cytotoxicity on vacc-BMLF1-infected and GLC peptide-loaded target cells. A selection of nine limiting dilution clones with the relevant specificity were used for TCR analysis along with the nine clones obtained by tetramer sorting, all of which proved to be GLC specific.

Table I gives details of TCR usage for all 18 primary clones, showing the V and J combinations and CDR3 amino acid sequences for both α- and β-chains of the receptor. Note that limiting dilution and tetramer-sorted clones showed similar distributions of TCR usage and therefore the two sets of data are combined in Table I. The most striking feature of the results is that three families of clonotypes dominate the primary GLC-specific response in this patient. The most abundant family (clones 259-3.1*) typically used an AV23S1/AJ3S2 α-chain rearrangement and conserved seven amino acid CDR3 sequences combined either with a BV16-positive, or less frequently a BV2-positive, β-chain with conserved nine amino acid CDR3 sequences. A second family (clones 158-1.5*) was totally dominated by one αβ-chain combination, namely, AV2S3/AJ17S10 coupled with BV13S3/BJ2S7, again with highly conserved CDR3 sequences in both chains. A third smaller family (clones 80-4.2*) all expressed the same BV1S1/BJ2S7 β-chain structure coupled with one of two α-chains.

The availability of a Vβ16-specific mAb allowed us to check whether the high frequency of BV16-positive clonotypes seen among primary IM74 clones in vitro was reflected in direct staining of the primary PBMC population. The results of triple staining with the Vβ16-specific mAb, a CD8-specific mAb, and the A2.01/GLC tetramer are shown in Fig. 1,A. Tetramer-reactive cells constituted 5.6% of the CD8⁺ T cell pool in the acute phase of the disease, and there was a coincident expansion of Vβ16-positive cells such that these now constituted 3.3% of the CD8⁺ pool; this represents a significant increase over the mean level of Vβ16 representation in the CD8⁺ pool of healthy control donors (0.69 ± 0.3%). More importantly, 23.4% of the tetramer-reactive population was Vβ16 positive, and in fact these cells constituted a substantial fraction (almost 40%) of the whole Vβ16-positive subset present in the blood at that time (Fig. 1,A, right panel). Much of the Vβ16 subset expansion within the circulating CD8⁺ T cells of IM74 was therefore attributable to the primary GLC-specific response. As an additional control, we repeated the triple stain analysis of IM74 PBMC now using mAbs specific for Vβ2 and Vβ22, subsets which in certain other patients do dominate primary GLC-specific responses (see later) but which were represented either sparsely (Vβ2) or not at all (Vβ22) among IM74-derived primary clones. Only 10% and 7% of tetramer-reactive cells in IM74 PBMC lay in the Vβ2 and Vβ22 subsets, and neither subset was significantly expanded (Fig. 1 A legend).

Donor IM74 was rebled 28 mo later, by which time the frequency of tetramer-staining cells had fallen to 0.2% of the CD8⁺ T cell population. In this case, in vitro stimulation of PBMC with the GLC peptide only yielded four T cell clones with epitope specificity when tested in cytotoxicity assays on GLC peptide-loaded and vacc-BMLF1-infected target cells. The results of their TCR analysis are shown in Table I; although numbers were small, none of these memory clones was related to any of the clonotype families that had dominated the primary response.

A second HLA-A2.01-positive patient studied in the acute phase of the disease, IM61, also showed a marked GLC-specific primary response, representing 4.4% of the expanded CD8⁺ population by tetramer staining. Here, an analysis of TCR usage among limiting dilution clones, detailed in Table II, showed that the bulk of the primary response was made up of two families of clonotypes, each of which was obviously related to the most abundant clonotype family seen in the primary response in IM74. Thus, the first IM61 family (clones 209-63) used an α-chain which, though it involved AV15S1 rather than AV23S1, was structurally very similar in the CDR3 region and identical in the AJ region to that of the dominant IM74 family. Furthermore, this α-chain was combined with a β-chain, involving BV2, which was closely related to one seen in that same IM74 family (cf Table I, clones 2.3* and 3.1*). The second family of IM61 clonotypes (clones 183-122) used a β-chain which was either identical or closely related to that of the dominant BV16-positive IM74 clones, in this case combined with one of three different α-chains. There were in addition several other unique clonotypes displayed by isolated clones within the primary GLC-specific response of IM61 (Table II, clones 190-157). The results of triple staining the primary PBMC population with the Vβ16-specific mAb, the CD8-specific mAb, and the A2.01/GLC tetramer (Fig. 1,B) again showed that a significant Vβ subset expansion in IM blood could be ascribed to a component of the epitope-specific response. Thus, tetramer-staining cells constituted 4.4% of the CD8⁺ T cell pool and 18.4% of these tetramer-staining cells were Vβ16 positive. This coincided with an expansion of the total Vβ16 subset (Fig. 1,B) to 2.0% of the total CD8⁺ pool (vs a control value of 0.69 ± 0.3%) and, as can be seen in Fig. 1 B (right panel), almost half of this enlarged Vβ16-positive population was made up of tetramer-staining cells.

Donor IM61 was rebled 40 mo after IM, by which time the tetramer-stained population had fallen to 0.2% circulating CD8⁺ T cells. GLC-specific memory T cells were isolated as before by limiting dilution of peptide-stimulated PBMC cultures; eight clones were identified as GLC epitope specific in cytotoxicity assays on GLC peptide-loaded and vacc-BMLF1-infected targets. Table II shows the results of their TCR analysis. Only one of these IM61 memory clones (c20) showed any relation to a dominant primary clonotype, namely, the use of an AV15S1/AJ3S2 α-chain which was almost identical to that of primary clones 209-63 but was now combined with a different β-chain; three other memory clones also used AV15S1 but without the characteristic CDR3 sequence described above. Interestingly, however, four memory clones 72-61 all displayed a characteristic AV3S1/AJ1S7 α-chain and BV13S1/BJ2S9 β-chain which was highly related to a clonotype represented once (c157) in the primary response (see Table II, italic sequences).

A third HLA-A2.01-positive IM patient, IM69, mounted a GLC-specific primary response which by tetramer staining constituted 4.0% of the expanded CD8 population. Here, the analysis of limiting dilution and tetramer-sorted clones revealed the presence of one dominant clonotype family (Table III, clones 95-7*) using AV15-positive α-chain and BV2-positive β-chain structures that were highly related to those dominant in the IM61 primary response (cf. Table II, clones 209, 243, and 63). In addition, we noted several unique clonotypes represented by single clones within the IM69 primary response (Table III, clones 72-26*). As shown in Fig. 1,C, triple staining of the original PBMC population confirmed that Vβ2-positive cells were numerically dominant (58.6%) within the tetramer-positive population. Again as a control, we extended the triple staining to include mAbs specific for Vβ16 and Vβ22, subsets which in other patients can dominate primary GLC-specific responses (Fig. 1, A and B, and see later) but which were represented either not at all (Vβ16) or only once (Vβ22) among IM69-derived primary clones. Only 1 and 9%, respectively, of tetramer-reactive cells in IM69 PBMC lay in the Vβ16 and Vβ22 subsets, and neither subset was significantly expanded (Fig. 1 C legend).

Donor IM69 was rebled 28 mo after IM when 0.9% of the circulating CD8⁺ T cell pool stained with the A2.01/GLC tetramer. In vitro reactivation yielded 14 memory clones that were functionally identified as GLC specific and their patterns of TCR usage are shown in Table III. As in donor IM61, only a single memory clone (clone 24) showed any relation to the dominant primary clonotype and here again this involved the AV15S1/AJ3S2 α-chain characteristic of that clonotype, now combined with a different β-chain BV22S1/BJ2S8. However, not all components of the primary response had disappeared (Table III, italic sequences). Thus, two individual clonotypes seen in the acute infection (clones 72 and 13*) were again detectable as single clonotypes with exactly the same α- and β-chain sequences in memory (clones 14 and 18). More interestingly, the dominant family of clonotypes found in memory (clones 23-25) used AV2S3/AJ16S5 and BV22S1/BJ2S8 rearrangements which were either identical or very closely related to a single clone identified in the primary response (clone 26*). Several other unique clones within memory (clones 14-30) also used AV2S3 but now combined to AJ1S3 and with a different pattern of CDR3 sequences; not all of these clones could be analyzed for Vβ usage and therefore they are only provisionally grouped as a second family of related memory clonotypes in Table III.

In the case of IM69, representative clones from the primary and memory GLC-specific repertoires could be expanded sufficiently to compare their fine specificities when tested against a set of peptide variants of the GLCTLVAML epitope sequence. Alanine replacement mutations were introduced individually at positions 1–6 and 8–9 of this sequence, and levels of variant peptide recognition were determined in cytotoxicity assays on peptide-loaded target cells. The results from individual clones are presented in Fig. 2, lysis of targets loaded with the variant peptides (shaded bars) being shown alongside that seen using the cognate peptide (black bar). Three clones representing the dominant clonotype family in the primary response (clones 104 and 7* with identical receptors, clone 1 with a related receptor) all showed strong selectivity for the cognate epitope sequence and only tolerated alanine substitution at position 3 and also, for clones 104 and 7*, at position 8. Interestingly, two other primary clones with unrelated TCR structures (clones 62 and 6*) showed a quite similar pattern of selective recognition. By contrast, another primary clone with a unique TCR structure, clone 8, showed a much broader pattern of recognition and could tolerate all alanine substitutions except those at positions 4 and 6 in the sequence. Among memory clones, such broad recognition of all variants except substitutions 4 and 6 was much more common. This is exemplified by the data from clones 23, 10, 22, and 31, each representing slightly different TCRs within the dominant family of memory clonotypes. A different clone (clone 9), representative of the second clonotype family found in IM69 memory, showed an even broader reactivity that now included recognition of the alanine substitution at position 6. Another clone with this second clonotype family, clone 14, had a very similar α-chain to clone 9 but a different β-chain. This clone was functionally distinct from clone 9 in its inability to tolerate substitutions at C-terminal positions 8 and 9, consistent with the crystallographic data on TCR/MHC alignment showing an interaction between the TCRβ-chain and the C-terminal region of the peptide in the MHC class I groove (3, 4). Viewing the results overall, only the Thr → Ala mutation at position 4 consistently abrogated T cell recognition by IM69 clones; since this mutant peptide satisfies the principle sequence demands for HLA-A2.01 binding (39), its loss of antigenicity probably reflects the importance of the Thr residue as a TCR contact point.

Cryopreserved PBMC from the memory bleeds of IM74, 61, and 69 were in very short supply; therefore, we could not analyze these populations by direct staining to determine how well the observed patterns of Vβ usage among in vitro-derived memory CTL clones (Tables I-III) reflected the Vβ composition of GLC-specific memory in vivo. We therefore chose to address this issue by studying the GLC-specific response in a long-term EBV carrier, donor CMc, who was A2.01-positive and who had shown evidence of CD8⁺ T cell reactivity with the A2.01/GLC tetramer in preliminary assays on PBMC. Using the same in vitro peptide stimulation protocol as employed earlier to analyze memory CTL responses in the post-IM donors, we established 17 T cell clones from donor CMc that were identified as GLC specific in cytotoxicity assays. Fig. 3,A presents their TCR sequence data; 12 of 17 clones used a common Vα15/Vβ4 receptor structure (which was identical to that seen in a single memory clone from IM 61: see Table II, clone 64), whereas 2 of 17 clones used a Vβ2 chain and 3 of 17 clones used a Vβ16 chain.

The recent development of a Vβ4-specific mAb, along with the existing reagents to other Vβ subsets, allowed us to compare the in vitro cloning data with the Vβ composition of donor CMc’s GLC-specific memory as determined by direct ex vivo staining. Cryopreserved PBMC from this donor were therefore triple stained with the anti-CD8 mAb, the A2.01/GLC tetramer, and individual mAbs specific for the relevant Vβ subsets. As shown in Fig. 3,B, 52.0% of the tetramer-staining cells in donor CMc memory lay within the Vβ4 subset; furthermore, smaller proportions of tetramer-positive cells were also detectable in the Vβ2 and Vβ16 subsets but were not found in several other subsets tested, of which Vβ22 serves as an example. The above distribution was observed in several independent blood samples taken from donor CMc over a 12-mo period. From the two sets of data presented in Fig. 3, it is clear that the pattern of Vβ usage within GLC-specific memory obtained using our in vitro cloning protocol is broadly consistent with that identified by direct ex vivo staining of PBMC; certainly in vitro cloning can identify the dominant Vβ component of memory.

Finally, to augment the earlier prospective studies on IM74, 61, and 69, we collected a new series of HLA-A2.01-positive IM patients and in each case followed the maturation of the GLC-specific response by direct ex vivo staining of PBMC. This work focused on the distribution of tetramer-positive cells among the Vβ2, Vβ16, and Vβ22 subsets since these subsets had each been identified as significant contributors to the GLC-specific response in the original in vitro cloning work (Tables I-III), and the relevant Vβ-specific mAbs were available; as a control we also included a Vβ18-specific mAb in the analysis, representing a Vβ-chain which had not previously been observed among in vitro-derived GLC-specific clones. Five IM patients (IM77, 78, 81, 83, and 96) were analyzed both in the acute phase of the disease, where 4–10% of the expanded CD8⁺ T cell pool in the blood stained with the A2.01/GLC tetramer, and between 12 and 33 mo later, by which time tetramer staining had fallen to 0.3–1.1% circulating CD8⁺ T cells. In the acute disease, all five patients showed marked skewing of the primary GLC-specific population into one or more of the Vβ2, Vβ16, or Vβ22 subsets, and this was again coincident with significant expansions in the size of the relevant Vβ subset in IM blood. However, when the same donors were rebled after IM, in three of five cases the Vβ subsets that had been dominant during primary infection were represented very poorly, if at all, in the tetramer-reactive CD8⁺ memory cell pool.

Illustrative data from one such patient, IM77, are shown in Fig. 4,A. Here, 55.2% of the primary tetramer-reactive population was Vβ22 positive, explaining why the Vβ22-positive subset had expanded to 6.3% of the total CD8⁺ T cell pool at this time (compared with a normal mean of 2.6% ± 1.0%, data not shown). By contrast, when this patient was studied 12 mo later, only 4.5% of the tetramer-reactive memory population was Vβ22 positive. A second smaller component of the primary GLC-specific response in IM77 lay within the Vβ2 subset and this was also essentially undetectable in the memory population, whereas a tetramer-positive Vβ16-positive component was detectable both in the primary response (as a subdominant component) and in memory. Fig. 4 B provides another such example where 49% of the primary tetramer-reactive population in IM patient IM81 was Vβ22 positive (contributing to an expansion of the Vβ22 subset to 8.5% of the total CD8⁺ T cell pool, data not shown) but 32 mo later only 5.3% of the tetramer-reactive memory population lay in this subset; a smaller component of the primary response had been Vβ2 positive, yet this later appeared to outnumber the Vβ22 component in memory. A third IM patient, IM83, likewise developed a highly amplified Vβ22-positive component constituting 58% of the primary tetramer-reactive population in the acute phase of the disease; 21 mo later, the representation of Vβ22-positive cells in the tetramer-stained memory population had fallen to 2% (data not shown).

We should however stress that not all patients displayed such dramatic change. Two individuals (IM78 and IM96) showed significant retention of the two major components of the primary epitope-specific response with the transition into memory. The point is illustrated in Fig. 4 C with reference to patient IM78. In this case, 48% and 29% of the primary tetramer-reactive population were Vβ2 and Vβ22 positive, respectively; these same components were clearly still represented in memory 33 mo later, though they now constituted only 10% and 8% of the total memory population. The primary responses in patient IM96 also contained significant Vβ2 and Vβ22-positive components which remained detectable 12 mo later in memory (data not shown).

In the present work, we have used the EBV model to study the natural history of a CD8⁺ T cell response in humans. Our first objective was to determine the clonal composition of the primary response to one of the immunodominant lytic cycle epitopes recently identified for this virus (34). This was important because our earlier finding of oligoclonal TCR usage within expanded Vβ subsets of IM CD8⁺ T cells had implied that such expansions were Ag driven (14), yet at the time no direct link could be made with responses to the then known virus-latent cycle epitopes. We reasoned that if some of the lytic Ag-specific responses involved focused TCR usage, then these more abundant reactivities (18) might contribute to the types of Vβ expansion described above. A recent study of TCRβ chain usage within primary CTLs specific for the B8-restricted RAK-lytic cycle epitope in fact offered no support to this idea; the 10 RAK-specific clones from three IM patients examined were all clonotypically unique (35). Here, we describe a more detailed analysis of the primary response to the A2-restricted GLC-lytic cycle epitope and show that this does indeed tend to be oligoclonal, that it involves a similar set of highly related clonotypes in different patients, and that it can influence Vβ subset representation in the CD8⁺ T cell pool.

Thus, in patient IM74 the bulk of the primary GLC-specific response was composed of three families of clonotypes, of which the dominant family carried an AV23S1 (or occasionally AV15S1) α-chain with a distinct CDR3 sequence and one of two β-chain structures, either BV16S1 or BV2S1, each with their own distinct CDR3 motifs (Table I). In patient IM61 the primary response was dominated by two clonotype families. One used the characteristic AV15S1 α-chain and BV2S1 β-chain structures described above for IM74; the other used the characteristic BV16S1 β-chain seen in IM74, now combined with a different α-chain (Table II). Finally, the same AV15S1/BV2S1 αβ-chain combination and associated CDR3 sequences seen in both the above patients was the dominant clonotype in the primary response of the third patient, IM69 (Table III). Interestingly, some of these same “public” clonotypes have also been identified within the GLC-specific T cell population found in the synovium of rheumatoid arthritis patients or in healthy donors with no recorded history of IM (40).

Further work went on to show that the patterns of clonal dominance apparent from in vitro cloning were consistent with the results from direct ex vivo staining of the primary GLC-specific response with the tetramer and with CD8 and Vβ subset-specific Abs. Thus, the dominant BV16 clonotypes seen among primary clones from IM74 and IM61 in both cases correlated with a significant concentration of tetramer-positive cells within the Vβ16 subset (Fig. 1, A and B). Likewise, the dominant BV2 clonotype among IM69 primary clones reflected the fact that more than half of the tetramer-positive cells in the original PBMC population were Vβ2 positive (Fig. 1,C). Such findings suggest that in vitro cloning has provided a representative picture of the in vivo response in these patients, at least in terms of identifying the more abundant clonotypes. Furthermore, in all three cases, the concentration of tetramer-positive cells within a particular Vβ subset was linked to a preferential expansion of that subset relative to the CD8⁺ T cell pool as a whole; later studies on other A2.01-positive IM patients (Fig. 4) provide further examples of this same phenomenon. This strengthens the view that many of the oligoclonal Vβ expansions that are frequently seen among CD8⁺ T cells in acute IM (14) actually represent EBV-driven responses, most likely against as yet unidentified lytic cycle epitopes.

In the context of these triple staining protocols, we noted that in some cases the profiles of tetramer vs Vβ staining appeared to show two dually stained populations of cells; one had high tetramer reactivity but relatively low Vβ staining, as might arise from competitive binding between reagents detecting the same molecular complex, whereas the other had lower tetramer reactivity and standard levels of Vβ staining (see for example Fig. 1, A and C). It remains to be seen whether these differences do reflect T cells with different affinities for the GLC tetramer. We believe however that the values quoted both here and in Fig. 4 for the size of tetramer-positive/Vβ-positive double-stained populations are not overestimates, even though they may include some cells with low tetramer staining. Thus, the quadrant boundaries for FACS analysis were always set carefully using multiple controls, the GLC-tetramer only produced staining above the relevant boundary when tested on cells from A2.01-positive EBV-infected individuals, and the IM-PBMCs being analyzed showed no staining above the boundary either with an A2.01/HIV epitope tetramer or with various tetramers carrying EBV epitope peptides complexed with other HLA alleles (data not shown).

The second major objective in the present work was to compare TCR usage within the epitope-specific primary response with that subsequently established in T cell memory. To date, this issue has been most closely examined in the context of the immunodominant latent cycle epitope FLR, where two independent studies demonstrated the persistence of dominant clonotypes from acute infection into memory (31, 32). However, primary responses to FLR are not only quantitatively less abundant than typical GLC responses but also are less dramatically culled with the development of memory. Thus, in our previous study, after allowing for the fall in total CD8⁺ pool size with recovery from IM, absolute numbers of circulating FLR-specific cells identifiable by tetramer-staining fell 5- to 20-fold between primary and memory bleeds (18); in contrast, the mean reduction in GLC-specific cell numbers in the eight patients studied here was 50- to100-fold. Although the detailed prospective analysis of TCR usage among GLC-specific clones was limited to three patients, it was interesting that in each case the clonotypes dominating the primary response were not detected in the in vitro-reactivated memory population; instead, there were new dominant clonotypes in memory that had only been detectable as unique subdominant clones within the primary response of the same individual. In the case of IM69, consistent differences in fine specificity were apparent between the dominant primary and memory clonotypes on testing against alanine-substituted versions of the cognate GLC sequence. Here, the memory clones were more tolerant of epitope sequence variation than primary clones (Fig. 2); further work will be required to see whether this is a general feature of maturing GLC-specific responses or simply a consequence of the particular clonotypes that come to dominate IM69 memory. Limitations on cell numbers prevented a detailed analysis of the affinity of these clones for the A2.01/GLC epitope, but initial peptide titration assays did not reveal obvious differences between clones (our unpublished observations). Recent work in murine models has provided interesting examples of epitopes where either CD4⁺ or CD8⁺ recall responses tend to recruit the higher affinity clonotypes from within a more heterogeneous primed population (40, 41, 42). In both of these models, however, maturation of the response was induced experimentally by a single re-exposure to Ag and involved only a slight narrowing of the primary TCR repertoire (42, 43). These circumstances are clearly different from the present study of a naturally persistent virus infection where, over a longer time period, one dominant clonotype appears to be replaced by another.

Again, from examples in other experimental systems (44, 45, 46, 47), it is possible that a reliance on in vitro-derived clones might have misrepresented the true content of functional epitope-specific memory in our donors, especially since additional aliquots of PBMC from the memory bleeds of IM74, IM61, and IM69 were not available for direct ex vivo analysis. Hence, the switch in dominant clonotype usage could be more apparent than real. Three points argue against this possibility. First, we used an optimized protocol for peptide-induced in vitro reactivation of GLC-specific memory (37) which, when applied to a healthy virus carrier with no history of IM, yielded a range of clonotypes that broadly concurred with the results of direct tetramer/Vβ double staining on PBMC ex vivo (Fig. 3). Second, after the initial in vitro stimulation with peptide, the cloning conditions used here were essentially the same as those used in our earlier prospective study of FLR epitope-specific responses during and up to 3 years after acute IM (32), a study in which the hierarchy of dominant clonotypes seen in individual donors remained constant over time. This strongly implies that in vitro cloning is capable of detecting a stable clonotype distribution where such a distribution is maintained in vivo. Third, and most important, extension of the work to five new A2.01-positive IM patients revealed further examples of changing patterns of TCR usage within the GLC-specific response over time. Here, the evidence came from the direct analysis of PBMC populations without clonal selection in vitro. In all five individuals the primary GLC-specific population, identified by tetramer and CD8 staining, showed highly skewed Vβ usage with dominant contributions from Vβ2-, Vβ16-, and/or Vβ22-positive cells. However, in three of five cases, the Vβ components that dominated the primary response were subsequently found to be represented very poorly, if at all, in the GLC-specific memory pool.

The overall results from clonotypic analysis and from Vβ subset representation indicate that, in a significant number of cases, clonotypes dominating the primary GLC-specific response in A2.01-positive IM patients do not remain dominant in epitope-specific memory, at least as this is visualized in the circulating CD8⁺ T cell pool. Such clonotypes may in fact be subject to a process of functional exhaustion leading to deletion like that reported for some epitope-specific responses in murine LCMV infection (17, 48). In that work functional exhaustion was ascribed to the effects of continued stimulation with high Ag load since it was best observed under conditions of overwhelming virus challenge with a persistent LCMV mutant (17). A recent analysis of CTL responses to primary HIV infection has also documented cases of apparent clonotype deletion following resolution of the acute symptoms, and this was likewise ascribed to the unusual capacity of HIV to persist at a very high viral load in vivo (23). The present paper suggests that this same phenomenon can occur in a situation where overall control of the viral infection is not lost. Thus, primary EBV infection is associated with high-level virus replication in the oropharynx and with the virus-driven proliferation of latently infected B cells throughout the lymphoid system. However, the virus load in both compartments is then promptly reduced, leading to a low-grade virus carrier state, which thereafter is kept under tight immunological control (26). It remains to be seen how frequently EBV-lytic cycle epitope responses follow the pattern described here and whether functional exhaustion/deletion of CD8⁺ clonotypes is a direct consequence of high antigenic challenge or a secondary consequence of an effect mediated at the level of CD4⁺ T cell help.

We are grateful to Dr. J. Faint (Department of Rheumatology, University of Birmingham) for help with FACS analysis and for details of Vβ subset distribution in healthy individuals and to Deborah Williams for excellent secretarial help.