Epitope Specificity and Clonality of EBV-Specific CTLs Used to Treat Posttransplant Lymphoproliferative Disease¹

Posttransplant lymphoproliferative disease (PTLD)³ is a potentially lethal complication of iatrogenic immunosuppression following solid organ and bone marrow transplantation (1, 2). PTLD occurs in up to 10% of all transplant recipients and in most cases is associated with EBV, a gammaherpes virus that infects around 90% of the adult population. Following primary EBV infection, normally during early childhood, the virus establishes a life-long asymptomatic infection in circulating B lymphocytes, which is effectively controlled by EBV-specific CTL (3, 4). In transplant recipients, however, immunosuppressive drug regimens suppress the function of EBV-specific CTLs and this reaction can lead to uncontrolled proliferation of EBV-infected B lymphocytes and tumor formation (5, 6). Reduction of immunosuppression, partially restoring the EBV-specific CTL function, is currently the first line treatment for PTLD and is successful in a proportion of cases (6). However, restoring immunity without inducing graft rejection is a fine balancing act, and tumor recurrences are common, often resulting in the emergence of more resistant clones. Further treatment strategies include radiotherapy, chemotherapy, surgery and anti-B cell mAb therapy (7, 8), but even with such secondary treatment options the mortality rate is around 50% (9, 10).

Our increased understanding of the mechanisms by which T lymphocytes recognize virus and tumor-specific Ags has led to a growing interest in new immunotherapeutic strategies. Adoptive T lymphocyte therapy is one such approach and involves the administration of CTL grown ex vivo to selectively reconstitute immunity in transplant recipients. T cell immunotherapy was first pioneered in the United States for the treatment of CMV infection in bone marrow transplantation patients in the early 1990s (11). Riddell et al. (11) demonstrated that a large number of virus-specific CD8⁺ memory T cells from bone marrow transplantation donors could be grown ex vivo, infused into recipients, and remain detectable for up to 1 mo postinfusion. Since then, adoptive transfer of autologous, ex vivo-grown, EBV-specific donor CTL has successfully prevented and treated EBV-positive PTLD in both bone marrow transplantation and solid organ transplantation patients (12, 13, 14, 15). However for general treatment purposes growing autologous CTL from individual donors or recipients is impractical and often not possible in the solid organ transplantation situation in which the organ donor is usually not available and the tumor usually arises from recipient’s cells. In addition, CTL may be difficult to grow from immunocompromised or EBV seronegative individuals and it can take several months to obtain a sufficient cell number for treatment, a time constraint that may be fatal for patients with rapidly growing tumors.

As an alternative to autologous CTL we recently investigated the use of partially HLA-matched allogeneic CTL for the treatment of EBV-positive PTLD. A frozen bank of 107 HLA-typed polyclonal CTL lines specific for EBV proteins was established from healthy blood donors (16). Donor CTL lines partly matched to the recipient on HLA-A, HLA-B, and HLA–DR alleles (a maximum of six allele matches) were tested ex vivo and the lines showing maximum specific and minimum nonspecific killing of recipient cells were chosen for infusion. The initial pilot study demonstrated complete remission of tumor in 3 of 5 PTLD patients treated and also highlighted the safe, cheap, and rapid use of the bank of allogeneic CTL (17). A subsequent phase II trial using the same bank of CTL further demonstrated the effective use of HLA-matched allogeneic CTL, with complete or partial response observed in 52% of patients 6 mo posttreatment. Significantly better tumor responses were seen in patients infused with CTL with a high degree of donor CTL recipient HLA that match and contain high proportions of CD4 T cells (18). During the phase II trial no attempt was made to characterize the epitope specificity of the infused CTL or to correlate this feature with patient outcome. However, the observed importance of donor CTL recipient HLA matching suggests that further refinement of this procedure may be beneficial. The present study was undertaken to investigate the HLA-restricted epitope specificity at the protein and peptide level of donor CTL used to treat 28 EBV-positive PTLD patients who completed the phase II trial, and to correlate the results with tumor response.

Establishment of the CTLs has been detailed elsewhere (16). Briefly, PBMC were obtained from HLA typed, EBV seropositive blood donors (Scottish National Blood Transfusion Service) and lymphoblastoid cell lines (LCL) and CTLs grown in culture medium as detailed in Wilkie et al. (18). CTLs were tested in standard chromium release cytotoxicity assays against PHA-stimulated blasts, autologous and HLA-mismatched LCL, and K562 cells. CTLs specific for autologous LCLs were analyzed by FACS for B cell, NK cell, and a wide range of T cell activation and differentiation markers and then frozen. CTL lines contained T cells, predominantly showing an activated CD8⁺ phenotype with a small proportion of CD4⁺ cells. A total of 24 of the infused CTLs were available for further analysis.

A cohort of EBV-associated PTLD patients was accrued as part of a phase II multicentre clinical trial detailed in Haque et al. (18). Patients from 19 transplant centers were recruited to the trial with informed written consent from patients or guardians. The study was approved by the Lothian Research Ethics Committee. The diagnosis of PTLD and its classification was determined by histological examination. Epstein-Barr-encoded small nonpolyadenylated RNA (EBER) staining was performed by in situ hybridization using a commercial kit from DAKOCytomation and immunohistochemistry conducted using commercial Abs for EBV-encoded nuclear Ag (EBNA)-1, EBNA-2, and latent membrane protein (LMP)-1 expression (DAKOCytomation). Assessment of tumor cell clonality was performed using a DAKOCytomation In Situ Hybridization kit for κ and λ mRNA. The 33 PTLD patients were treated with CTLs from the frozen bank on a HLA-matched basis and monitored at regular intervals for tumor regression: 28 patients completed the trial. Three CTL lines were each used to treat more than one patient and one patient was given two lines due to insufficient cell numbers. A positive response (responder group, complete or partial response; n = 17) was defined as disappearance of, or at least a 50% reduction in, tumor mass with stabilized or improved clinical condition, whereas a negative response (nonresponder group, n = 11) was defined as stable or increased tumor mass or deterioration in clinical condition. Patient details are listed in Table I.

Monocytes were isolated from donor PBMC samples via a negative selection method using the Monocyte Isolation kit II (Miltenyi Biotec) as per the manufacturer’s instructions and the purity estimated with anti-CD14PE Ab (BD Pharmingen) and FACS analysis. Isolated monocytes were then resuspended in 10 ml of culture medium (RPMI 1640 medium containing 10% FBS, 4 mM l-glutamine, 100 IU/ml penicillin-streptomycin, 25 mM HEPES, 20 ng/μl GM-CSF (R&D Systems), and 20 ng/μl IL-4 (R&D Systems) and incubated for 7 days at 37°C with 5% carbon dioxide. Three ml of fresh culture medium was added on days 3 and 6, and on day 7 the cells were harvested. DC purity was assessed with anti-CD14PE and DC-SIGN FITC (R&D Systems) Abs and FACS analysis.

A total of 1 × 10⁴ DCs were infected with recombinant vaccinia virus (multiplicity of infection 10:1) constructs provided by Professor A. B. Rickinson (Institute for Cancer Studies, Birmingham, U.K.) expressing one of the following EBV Ags: EBNA-1, EBNA-2, EBNA-3A, EBNA-3B, EBNA-3C, EBNA leader protein, LMP-1 and LMP-2. A recombinant vaccinia virus construct with no EBV Ag insertion was included as a control. A known number of infected DCs was assessed for RNA transcription of specific protein by RT-PCR and compared with a similar number of LCL cells. All infected DCs showed comparable transcription levels (see supplemental Fig. 1).⁴ Infected DCs were incubated with ⁵¹Cr (10 μCi) for 1 h at 37°C, washed, and used as target cells in the assay. Target cells were incubated with 1 × 10⁵ CTL (10:1 E:T ratio) for 4 h at 37°C, centrifuged and 100 μl of supernatant removed for analysis of chromium release in the 1480 Wizard 3″ Automatic Gamma Counter (PerkinElmer). Incubations were performed in triplicate and the average count taken to estimate the percentage of specific cell lysis. A response was deemed positive if the percentage of specific lysis was greater than that obtained with the empty vaccinia virus control. The highest percentage of specific lysis obtained was termed a dominant response, whereas lower percentages were termed subdominant.

ELISPOT assays were conducted using the Human Interferon-Gamma kit from R&D Systems as per the manufacturer’s instructions. Briefly, plates were equilibrated with culture medium before addition of cells. CTLs (5 × 10⁴) were added in combination with either 1 × 10⁴ autologous LCLs or 1 × 10⁵ autologous PBMC or 1 × 10⁵ autologous PBMC with EBV specific peptide (10 μg/ml) (Table II). The 100 μl of prepared cell combinations were added to triplicate wells and incubated at 37°C with 5% CO₂ for 24 h. IFN-γ-producing CTLs were then detected with the kit detection system and counted in an ELISPOT reader (Autoimmun Diagnostika). Triplicate counts were averaged to give the final count.

RNA was extracted from 5 × 10⁶ CTLs using the RNeasy mini kit and QiaShredder system (Qiagen) as per the manufacturer’s instructions. DNase (Promega) treatment was performed on 1–2 μg of extracted RNA, and 1 μg of DNase treated RNA was converted to cDNA using random hexamers and the Thermoscript RT system (Invitrogen). Functionally rearranged TCR β-chain variable gene subfamilies were amplified across the CDR3 encoding regions using 23 subfamily-specific primers and a FAM-conjugated β-chain constant region specific primer (19, 20). PCR amplifications were performed on 1 μl of cDNA in a total volume of 20 μl containing 10 mM Tris-HCL (pH 8.3), 50 mM KCl, 2 mM MgCl₂, 200 μM each of dATP, dTTP, dGTP, dCTP, 1 μM variable and constant primers, and 0.5 U of Amplitaq Gold polymerase (Applied Biosystems). Cycling conditions were 95°C for 10 min followed by 30 cycles of 95°C for 30 s, 58°C for 30 s, 72°C for 45 s, and a final extension at 72°C for 5 min. One microliter of PCR product was diluted in 10 μl of nuclease free water and then further diluted 1/10 with Hi-Di formamide (containing Genescan 500LIZ size standard) before electrophoresis in an ABI 3730 (dye set 5) automated sequencer. ABI GeneMapper software (version 3.7) was used to analyze data.

Comparison of CTL protein recognition and polyclonal or clonal TCR usage were compared between treatment responder and nonresponder groups using χ² (3 × 2) contingency tables. The Fisher’s Exact test was used to analyze CTL latent protein specificity across treatment groups.

A modified chromium release assay was used to determine which of the eight EBV latent proteins were preferentially recognized by 21 of 24 CTL lines (three lines were not tested due to insufficient cell numbers). Peripheral blood DCs from blood donors were infected with recombinant vaccinia virus constructs containing one of the following EBV latent proteins: EBNA-1, -2, -3A, -3B, -3C, EBNA leader protein, and LMP-1, LMP-2. Infected DCs were used as targets in the assay. The autologous LCL and an empty vaccinia virus construct were included as controls. EBV latent protein recognition by CTL lines was polyspecific in nature. Fig. 1,A displays the results obtained for one line showing a dominant response to EBNA-3C and a subdominant response to LMP-1. Of the 21 CTL lines tested, four (19%) produced either a response against one EBV protein, nine (43%) produced a response against two EBV proteins, and eight (38%) produced a response against three or more EBV proteins. Analysis of dominant responses alone revealed that CTL lines recognized mainly EBNA-3 proteins with 43% of lines producing a response against EBNA-3C, 24% against EBNA-3A, and 19% against EBNA-3B (Fig. 1 B). When both dominant and subdominant responses were analyzed the majority (95%) of the CTL lines tested recognized at least one of the EBNA-3 proteins with 28% recognizing EBNA-3A, 28% recognizing EBNA-3B, and 71% recognizing EBNA-3C. Subdominant responses were seen in a number of CTL lines against the LMP proteins with 24% recognizing LMP-1 and 33% recognizing LMP-2. Subdominant responses were observed in a small number of lines against the EBNA-1 (19%), EBNA-2 (9%), and EBNA leader proteins (24%), with one line displaying a dominant response toward EBNA-1 (70%, EBNA-1-specific lysis).

The chromium release assay using recombinant vaccinia vectors can be insensitive and therefore may not have detected low frequency T cell populations. Therefore, human IFN-γ ELISPOT assays were used to confirm the protein specificity at the peptide level in 20 of 24 CTL lines (PBMC numbers were insufficient in four lines). CTLs were tested against a variety of peptides with the same HLA-A or HLA-B restriction as detailed in Table II, and a control peptide with a mismatched HLA allele restriction. LCLs were also included as a positive control for IFN-γ release. A total of 17 CTL lines displayed a positive response toward peptides with the same HLA subtype restriction and to those peptides derived from proteins corresponding to CTL protein specificity (Table III). In the case of eight CTL lines (lines 7, 9, 44, 47, 50, 67, 68, and 91) more than one HLA-restriction was used (Table III). Four CTL lines (8, 48, 58 and 86) of the 17 CTL lines recognized peptides derived from one EBV protein, whereas the remaining 13 CTL lines recognized peptides derived from two or more proteins. Fifteen CTL lines recognized at least one latent EBV Ag and 12 CTL lines recognized one of two lytic EBV Ags (BRLF1 or BZLF1). The remaining CTL lines (three lines in total) did not respond to the panel of peptides.

A total of 28 PTLD patients were treated with CTL and monitored for tumor regression 6 mo posttreatment. Data were available for all CTL lines used to treat the 11 nonresponders and for CTL lines from 14 of the 17 responders. Within the responder group of 14, three (21%) of CTL lines recognized one EBV protein, eight (57%) recognized two EBV proteins, and three (21%) recognized three or more EBV proteins. No significant difference was noted within the nonresponder group with 1 of 11 nonresponders (9%) of CTL lines recognizing one EBV protein, 4 of 11 (36%) recognizing two EBV proteins, and 6 (54%) recognizing three or more EBV proteins (p = 0.22). Analysis of the specific proteins recognized showed that the majority of CTL lines from both the responder and the nonresponder groups recognized the EBNA-3C protein (64% and 73%, respectively). The number of CTL specific for EBNA-3A and EBNA-3B proteins was slightly increased in the responder group compared with the nonresponder group (EBNA-3A, 36% responders, 18% nonresponders; EBNA-3B, 36% responders and 18% nonresponders), but not significantly so (p = 0.4 for both EBNA-3A and EBNA-3B). In comparison, the number of CTL displaying a response to EBNA-2 and EBNA leader protein was decreased in the responder group compared with the nonresponder group (EBNA-2, 7% responders, 27% nonresponders; EBNA leader protein, 14% responders and 36% nonresponders), but again not significantly so (p = 0.28 and p = 0.35, respectively). CTL responses to LMP-1, LMP-2, and EBNA-1 were similar in both treatment groups (LMP-1, 21% responders and 27% nonresponders; LMP-2, 29% responders and 36% nonresponders; EBNA-1, 14% responders and 18% nonresponders) (Fig. 2).

Where available CTL specificity at the peptide level was compared between treatment responders (n = 14) and nonresponders (n = 7). Within the nonresponder group, infused CTL peptide specificity was restricted to one, two, or three EBV proteins (14%, 14%, and 72% of nonresponders, respectively) and one or two HLA alleles. In two cases (patients treated with CTL 9 and CTL 18) the restriction element of the CTL (CTL 9, A2 or B7; CTL 18, B8) was not shared with the recipient. A similar recognition of proteins and HLA alleles was observed for the responder group; however, in some cases as many as five EBV proteins were recognized by CTL (one protein, 29%; two proteins, 14%; three proteins, 36%; four proteins, 7%, and five proteins, 14%) and up to three HLA restrictions (CTL 67). In all responder cases, the restriction element of the CTL matched to at least one recipient HLA allele (Table III). Interestingly, the matching HLA restriction elements in the nonresponder group were either A2, A11, or B7, whereas the matching HLA restriction elements in the responder group were A2, B7, B8, B35, B44, or B62 with 6 of the 14 responders (43%) matching on B8 (Table III).

Although the majority of PTLD cases showed the classic histological features of the disease (classified as hyperplastic (n = 4), polymorphic (n = 8), or monomorphic (n = 10) type), five patients with Hodgkin’s-type PTLD and one patient with a Burkitt’s-like PTLD were included in the trial. These tumors are known to have a restricted pattern of viral gene expression that is less likely to be recognized by CTLs selected only on the basis of the best HLA match (21, 22, 23). We therefore analyzed viral gene expression in tumor biopsy specimens and protein specificity of infused CTLs and compared this with tumor response in these six patients.

All six tumors were EBER-positive, and five of the five Hodgkin’s-like PTLD cases expressed monoclonal κ or λ mRNA. However, the Burkitt’s-like PTLD patient stained for neither κ nor λ mRNA. Of the four Hodgkin’s-type PTLD cases with sufficient tumor material to test, all stained LMP-1-positive and EBNA-2-negative indicating a latency type II phenotype (Fig. 3). The Burkitt’s-like PTLD was negative for both markers, suggesting a latency type I gene expression (Table IV) (Fig. 3). Because there are no available Abs to the EBNA-3 proteins to stain formalin-fixed material we attempted to extract total RNA from tumor tissue sections for RT-PCR analysis. However, this attempt failed to provide amplifiable RNA.

At 6 mo after the last infusion, four of the five Hodgkin’s-type PTLD cases, and the one Burkitt’s-like tumor, showed a complete response to CTL therapy, whereas the remaining Hodgkin’s-type tumor showed no response. Based upon the chromium release assay data, all six recipients were treated with CTL lines with a dominant EBNA-3 specificity: four tumors received CTL lines with multiple protein specificity and two received CTL lines specific for EBNA-3A alone. One line also expressed subdominant reactivity against the latency type II proteins LMP-1 and LMP-2, and two CTL lines against the EBNA-1 protein. Analysis using data obtained from ELISPOT assays showed further reactivity against the lytic proteins BRLF1 and BZLF1, which were recognized by four CTL lines, and LMP-2 (recognized by two CTL lines) (Table IV). All six patients were treated with a mixed CD8/CD4⁺ population of CTL. Of the six recipients treated wit CTL lines, four recipients had less than 5%, one had 7%, and one had 26% CD4⁺ T cells.

CTL lines (n = 24) were analyzed for use of the TCR β-chain variable gene subfamilies using a spectratyping PCR. Subfamily usage was designated according to the number of specific peaks observed: three or more peaks in a Gaussian or skewed-Gaussian distribution were termed polyclonal usage, a single or double peak was termed clonal. Polyclonal and clonal peaks were observed for 22 of the 23 subfamilies; no polyclonal peaks were seen for subfamily 25. No one particular subfamily was preferentially used by the CTL lines. However, a polyclonal distribution pattern was particularly enhanced for subfamily 9 (polyclonal: 79% (19/24); clonal: 17% (4/24); undetected: 4% (1/24)) (Fig. 4).

CTL TCR spectratyping data were available for 16 treatment responders and 11 nonresponders. A significant difference was observed between the responder and nonresponder groups for subfamily 2 with 12 of the 16 (75%) responders and 3 of the 11 (27%) nonresponders presenting a polyclonal distribution, 3 of 16 (19%) responders and 8 of 11 (73%) nonresponders presenting a clonal distribution, and undetectable usage in 1 of 16 (6%) responders (p = 0.01). Subfamily 3 and subfamily 9 also had an increase in polyclonal distribution within the responder group compared with the nonresponder group (subfamily 3, 13 of 16 (81%) responders and 5 of 11 (45%) nonresponders, p = 0.05; subfamily 9, 15 of 16 (94%) responders and 6 of 11 (54%) nonresponders, p = 0.05) (Table V) (Fig. 5). A reduced polyclonal distribution for subfamily 5 and subfamily 16 was observed in the responder group compared with the nonresponder group, but this reduction did not reach significance (subfamily 5, 6 of 16 (37%) responders and 7 of 11 (64%) nonresponders, p = 0.4; subfamily 16, 3 of 16 (19%) responders and 5 of 11 (45%) nonresponders, p = 0.24) (Table V).

T cell immunotherapy for cancer is a fast expanding field, particularly for virus-associated tumors in which clinical trials using T cells targeting viral Ags have demonstrated safety and efficacy (24, 25, 26, 27). Furthermore, recent trials for melanoma treatment using the same approach and targeting tumor-specific Ags have produced encouraging results (28). However, immunotherapy-based treatments remain problematic for those cancers for which viral- or tumor-specific Ags have yet to be identified.

In a recently reported phase II clinical trial using allogeneic, EBV-specific CTL lines to treat PTLD, we have shown that tumor response significantly increased with the number of CTL-recipient HLA allele matches (varying from two to six matches) and the percentage of CD4⁺ T cells in the infused CTL (18). In the present study, we sought further correlates of tumor response in the trial participants that might be important in future trials by characterizing the epitope specificity and clonality of the infused CTL lines.

Analysis of CTL protein specificity revealed that the majority (81%) were directed against two or more EBV latent proteins, and this protein specificity was confirmed at the peptide level (HLA-restricted) in 17 of the CTL lines. In line with previous studies on ex vivo grown, EBV specific CTL, the lines predominantly recognized the EBNA-3 proteins, in particular EBNA-3C (4, 29). Because most PTLDs display full latent viral gene expression, CTLs with specificity for the EBNA-3 proteins would be expected to recognize and kill the tumor cells. However, when CTLs were chosen for specific recipients on HLA allele matching alone, without knowledge of the peptide specificity and functional HLA restriction of the CTL, mismatches clearly occurred (see Table III, where recipients matched with the CTL donor 9 and donor 18 did not match with the predominant peptide specificity of the CTL that was restricted through HLA-A2, B7 for CTL 9, and B8 for CTL 18). Thus we speculate that prior knowledge of CTL specificity and HLA restriction at the peptide level may have enhanced the CTL-recipient matching process and consequently be expected to improve tumor response and patient survival.

Investigation of CTL specificity directed against lytic EBV peptides revealed that 70% of the lines also had specificity for lytic EBV Ags. The presence of EBV lytic reactivity in CTL lines had no bearing on patient response with 71% of nonresponders and 78% of responders receiving a CTL line with lytic reactivity. This finding is perhaps not surprising as lytic EBV Ags are rarely expressed by EBV-associated tumors in vivo; this finding is especially so for the more Ag-restricted tumor types. Therefore, the presence of lytic reactivity within the CTL lines is unlikely to be responsible for the tumor regression observed in this cohort.

The complete resolution of Hodgkin’s-type and Burkitt’s-like PTLD tumors in five of six cases, which was sustained at 6 mo, was unexpected. In general these tumors can be more difficult to treat when using conventional therapy. Moreover, they were all treated with CTL lines with a predominant specificity for EBNA-3 proteins; proteins not generally expressed by these tumor types. With limited formalin-fixed material available, we demonstrated a latency type II (Hodgkin’s-type cases) phenotype by the absence of EBNA-2 expression in tumor cells, and a latency type I (Burkitt’s-like case) phenotype by the absence of EBNA-2 and LMP-1 expression. However, rare Burkitt’s cell lines with mutated EBNA-2 genes that express EBNA-3A, EBNA-3B, EBNA-3C, and EBNA leader protein in the absence of EBNA-2 have been reported (30). Extrapolation of these findings to the in vivo situation could account for the response of the Burkitt’s-like tumor, but to our knowledge no such mutants have been identified in vivo.

Tumor cell killing in two of the five successful EBV Ag-restricted cases (recipients 1 and 4; Table IV) could have been mediated by subdominant clones within the CTL with HLA class I-restricted specificity for EBNA-1, LMP-1, and LMP-2, but in the other cases no such activity could be detected. Similarly, HLA-matched T cell populations with frequency below the limit of assay detection may also have contributed to the observed response; these populations could in theory expand after infusion and elicit an antitumor response. However, the banked allogeneic CTL lines have been stimulated by LCL and grown in vitro for ∼12 wk before freezing and attempts to continue in vitro culture from frozen lines had limited success, presumably due to their terminally differentiated and highly activated state (16). Furthermore, in our phase II trial, detection of infused allogeneic CTL in vivo by CTL precursor and TCR spectratyping assays for a small subset of patients showed no evidence of CTL expansion (18); however, these assays may not have detected lower frequency T cell populations and therefore T cell expansion following infusion cannot be ruled out.

EBV Ags, in particular EBNA-1, LMP-1 and LMP-2, have all been identified as a targets for HLA class II-restricted, CD4⁺ T cell-mediated killing (31, 32) and could have therefore formed a suitable target for the small CD4⁺ T cell population in the infused CTL if HLA class II matched to the recipient. However, this mechanism was not assessed in this particular study. The majority of CTL lines infused in both the responder and nonresponder patient groups were matched on the HLA-DR locus (82% and 73%, respectively), therefore HLA-DR matching alone was not a significant factor in terms of tumor response. This determination was clearly exemplified in the Hodgkin’s-type and Burkitt’s-like cases in which of the two EBNA-1-specific CTL lines used, only one was matched on the HLA-DR locus. This one was the nonresponder Hodgkin’s–type case (recipient 3, Table IV). Further possible matching with HLA-DQ and HLA–DP class II alleles was not assessed in this study but may also provide further specificity.

Although HLA-restricted protein and peptide specificity of CTL was clearly important for patient outcome in a number of cases (Table III), and a significant relationship between HLA matching and a favorable patient outcome was observed in our previous clinical study, it is also possible that the infused allogeneic T cells in the Burkitt’s-like and Hodgkin’s-type cases initiated an inflammatory response, which either reactivated endogenous low-level EBV-specific CTL or recruited nonspecific cytotoxic cells to the tumor site. Such a response would be similar to the beneficial graft-vs-leukemia response often seen following allogeneic bone marrow transplant in which donor lymphocytes display antitumor effects primarily through T cell recognition of mismatched minor histocompatability and tumor-associated Ags (33). The small CD4 T cell population within some of our lines (Table IV) may contribute such an effect; although, the remit of this study did not allow for detection of any such responses. However, recognition of possible antitumor targets by CD4 T cells has been reported (34) with CD4 T cells found to recognize autologous LCL but not EBV-negative B lymphoblasts, or, EBV-specific latent and lytic Ags, suggesting that some CD4 T cells may recognize a tumor-specific or an as yet unidentified EBV-specific Ag.

TCR spectratyping analysis of the CTL lines showed that no single TCR subfamily was preferentially used. This result was not unexpected when considering the polyclonal nature of the lines, the fact that the lines were stimulated weekly with IL-2 and LCLs expressing all the latent viral Ags, and that they contained a mixed population of CD8⁺ and CD4⁺ T lymphocytes (CD4⁺ T cell percentage range: <1–60%) (18). However, it is interesting that for some TCR subfamilies (notably 2, 3, and 9) a polyclonal distribution was significantly more likely to induce a tumor response than a monoclonal distribution. This result was probably because of the wider spectrum of epitope specificities inherent in the polyclonal distribution, and indeed, this response was the reason why no attempt was made to clone the banked CTL lines grown for in vivo use.

In summary, the results of this study suggest that, in conjunction with donor and recipient HLA allele matching, mapping CTL peptide epitope specificity before CTL infusions would enhance patient responses by identifying those epitopes restricted through the recipient HLA alleles. These improved CTL selection criteria may now allow other EBV-associated tumors with restricted EBV latent gene expression to be treated more effectively with CTLs. However, because we found no relationship between CTL protein specificity, tumor EBV Ag expression, and outcome in Hodgkin’s-type and Burkitt’s-like PTLD cases, other factors, such as infusion or expansion of undetectable minor EBV-specific HLA-matched T cell clones or activation of nonspecific cellular responses may be responsible for tumor regression in these cases. Such observations may enhance ongoing or future allogeneic-based T cell therapies directed against viral-associated tumors; in particular regimes using multivirus specific CTL in which the range of epitopes recognized by the CTL may be limited via antigenic competition during the CTL culture (35, 36).

We thank Gwen M. Wilkie, Marie M. Jones, Phoebe Wingate, and David Burns for coordination of the phase II trial, establishment, and phenotype analysis of the CTL bank, and processing of patient samples. We also thank Professor A. B. Rickinson for supplying many of the peptides used in this study.

The authors have no financial conflict of interest.