Spontaneous T cell responses to Epstein-Barr virus-encoded BARF1 protein and derived peptides in patients with nasopharyngeal carcinoma: Bases for improved immunotherapy

Abstract

Immunotherapy approaches targeting Epstein-Barr virus (EBV)-encoded antigens induce objective clinical responses only in a fraction of patients with undifferentiated nasopharyngeal carcinoma (UNPC). In the present study, we have characterized the immunogenicity of the EBV-encoded BARF1 oncogene with the aim to assess whether this protein could constitute a new target antigen for immunotherapy in this setting. Spontaneous CD4+ and CD8+ T cell responses specific for the recombinant p29 BARF1 protein were detected by IFNγ-ELISPOT in both EBV-seropositive donors and UNPC patients, but not in EBV-seronegative individuals. Using immunoinformatic prediction tools, we have selected 5 different candidate BARF1 T cell epitopes presented by HLA-A*0201. Although only one of these peptides was able to bind HLA-A2 with low affinity in the T2 stabilization assay, all 5 BARF1 nonamers readily elicited specific CD8+ T cell responses in EBV-seropositive HLA-A*0201+ donors and UNPC patients. Notably, the magnitude of CD8+ T cell responses to the whole BARF1 protein and derived A*0201 peptides was significantly higher in UNPC patients than in healthy donors. Moreover, cytotoxic T lymphocytes specific for the p2–10, p23–31, or p49–57 BARF1 peptides were easily obtained from HLA-A*0201+ donors. These cultures were not only able to lyse autologous targets loaded with the antigenic peptide, but also recognized tumor cells endogenously expressing BARF1 in an antigen-specific and HLA-A2-restricted manner. These findings, indicate that BARF1 is a particularly attractive antigen with immunogenic properties in most UNPC patients and provide valuable information to develop new strategies to improve the efficacy of EBV-targeting immunotherapy of UNPC patients. © 2008 Wiley-Liss, Inc.

Undifferentiated nasopharyngeal carcinoma (UNPC) is an epithelial tumor characterized by peculiar epidemiologic and clinic-pathologic features.1, 2 Although the incidence of UNPC is low throughout most of the world, this malignancy is endemic in a few, well-defined areas, particularly in Southern China and Southeast Asia. Multiple factors including genetic, environmental and Epstein-Barr virus (EBV) infection likely contribute to the complex pathogenesis of UNPC. In particular, the strong association with EBV is supported by the observation that, regardless of geographic origin, EBV genomes are detected in virtually all UNPC and are monoclonal.1, 2 Moreover, high titers of anti-EBV antibodies, especially of IgA, characterize UNPC patients at diagnosis and may precede tumor development by several years.1 EBV infection in UNPC cells is characterized by the expression of a limited set of latent proteins, including EBNA-1, LMP-1 and LMP-2, which have transforming properties and may serve as target for therapy.1, 2 These viral antigens, however, behave as weak immunogens, being able to provide only subdominant epitopes for cytotoxic T lymphocytes (CTLs).3 The ability of UNPC cells to process and present endogenously synthesized proteins to HLA class I restricted CTLs4, 5 has been recently exploited by anti-EBV immunotherapy strategies aimed at improving the clinical control of UNPC.6–8 The results of the initial trials are encouraging since the infusion of EBV-targeted autologous CTLs was shown to enhance specific immune responses and to induce objective clinical responses, although in a proportion of UNPC cases.7, 8 On these grounds, there is the need to devise new strategies aimed at boosting EBV-specific CTL responses to improve the extent and duration of clinical responses in UNPC patients.

Recent evidence indicates that UNPC frequently express BARF1,9, 10 an EBV-encoded protein of 221 amino acid able to induce oncogenic transformation of rodent fibroblasts or human B lymphoma cell lines11 and immortalize monkey kidney primary epithelial cells.12 While BARF1-immortalized epithelial cells are not tumorigenic in nude mice, ectopic expression of BARF1 in immortalized cell lines allows these cells to grow in vivo, suggesting that BARF1 may have a role in both immortalization and malignant transformation.13–15 BARF1 protein may be also secreted after signal sequence (residues 1–20) cleavage,15, 16 thereby promoting cell proliferation in a paracrine fashion. Moreover, BARF1 protein may function as a soluble receptor for human colony-stimulating factor-115 and could regulate immune response by inhibiting α-interferon secretion by mononuclear cells.17 Notably, BARF1 is also able to induce humoral responses in EBV-seropositive individuals and may serve as a target for antibody-dependent cellular cytotoxicity in UNPC patients.18 Nevertheless, no information is available on the ability of BARF1 to induce T cell responses in healthy donors or in UNPC patients.

In the present study, we investigated whether BARF1 protein is a target for spontaneous T cell recognition in healthy individuals and UNPC patients from Italy. We also aimed at identifying and validating HLA-A*0201-restricted BARF1 CTL epitopes. This with the final goal to provide the rationale to improve the efficacy of EBV-targeted immunotherapy of UNPC by boosting BARF1-specific T cell responses.

Material and methods

UNPC patients and healthy donors

Eighteen tumor samples, diagnosed according to histological criteria provided by W.H.O. classification,19 10 biopsies of normal nasopharyngeal mucosa, and 10 blood samples were obtained from patients with UNPC. All UNPC cases investigated were EBV-associated as shown by in situ hybridization for EBERs. Buffy coats from 13 healthy donors (11 EBV-seropositive and 2 EBV-seronegative) were also collected and included in the present study. Peripheral blood mononucleated cells (PBMCs) were isolated on Ficoll density gradients and cryopreserved immediately. All specimens were obtained after informed consent from both patients and donors. HLA-A and -B typing was performed in all cases by sequence-based typing, according to standard high-resolution typing techniques.

Cell lines and culture conditions

The following human cell lines were used in this study: the Akata and Raji Burkitt's lymphoma cell lines, the Granta 519 mantle cell lymphoma cell line (HLA-A2), the DAA3 EBV-transformed B lymphoblastoid cell line, and the transporter associated with antigen-processing-deficient T2 cells transfected with the HLA-A*0201 gene (T2-A2). EBV-transformed LCLs were generated in vitro by transformation of B cells using the standard EBV isolate B.95.8. All cell lines were cultured in RPMI-1640 (Gibco, Grand Island, NY), containing 10% fetal bovine serum (Gibco), 2 mM L-glutamine, 100 μg/ml streptomycin and 100 IU/ml penicillin (Sigma), with the exception of Granta cell line, which was cultured in complete Dulbecco's Modified Eagle's Medium (DMEM, Cambrex Bio Science Walkersville, MD). Induction of the EBV lytic cycle was achieved by incubation of cells with a combination of 12-O-tetradecanoyl-phorbol-1-acetate (Sigma, 20 ng/ml final concentration) and sodium butyrate (Sigma, 3 mM final concentration) in standard medium for 48 hr.

Analysis of BARF1 mRNA expression

Total RNA was extracted from cell lines and UNPC tumor tissues with Trizol reagent (Invitrogen, Carlsbad, CA), and then treated with DNase I (Promega, Milan, Italy). Reverse transcriptase PCR (RT-PCR) was carried out as described.10 Briefly, 1 μg of total RNA was used for a first-strand cDNA synthesis in a 20 μl reaction volume using 0.5 μg of random primers. Reverse-transcription was done with AMV-reverse transcriptase (Promega) according to manufacturer's instructions. cDNA aliquots were then subjected to PCR analysis using primer pairs specific for BARF1 and glycealdehyde-3-phosphate dehydrogenase (GAPDH) (Sigma, Milan, Italy). Sequences of primer pairs were as follows, BARF1: 5′-GGCTGTCACCGCTTTCTTGG-3′ and 5′-AGGTGTTGGCACTTCTGTGG-3′; GAPDH: 5′-GCCTCCTGCACCACCAACTG-3′ and 5′-CGACGCCTGCTTCACCACCTTCT-3′. Amplification conditions were: denaturation step at 94°C for 3 min, annealing at 60°C for 1 min, extension at 72°C for 1 min for 30 cycles and a final extension step at 72°C for 5 min.

Peptide-prediction analysis, bioinformatics and peptide synthesis

The BARF1 protein sequence spanning the first 60 amino acids and including the conserved transforming domain20 was analyzed by immunoinformatic tools to identify candidate antigenic epitopes presented by the HLA-A*0201 allele. Peptides were selected on the basis of their predicted binding affinity to HLA-A*0201 according to SYFPEITHI (www.syfpeithi.de) and BIMAS (www-bimas.cit.nih.gov/molbio/hla_bind/) prediction softwares available on the web. The HLA-A*0201-restricted Flu matrix 1 (M1_58–66) peptide (GILGFVFTL) was used as positive control. All peptides were synthesized by fluorenylmethoxycarbonil synthesis (Primm, Milan, Italy) and purity (>95%) was determined by reverse-phase high-performance liquid chromatography and verified by mass spectral MALDI-TOF analysis. Peptides were dissolved in DMSO at a concentration of 2.5 mg/ml and stored at −70°C until use. Work stocks for each peptide were prepared in phosphate-buffered saline at a final concentration of 500 μg/ml and stored frozen.

MHC stabilization assay

Selected peptides were tested for their ability to bind to HLA-A*0201 molecules in a MHC-class I stabilization assay using the T2-A2 cell line. Briefly, T2-A2 cells (2 × 10⁵) were pulsed with 20 μg/ml peptide and 5 nM β₂-microglobulin (Chemicon, Milan, Italy), then incubated at 26°C for 16 hr, followed by 3 hr at 37°C. HLA-A*0201 expression was then measured using the mouse anti-human HLA-A2, -A28 PE-conjugated BB7.2 monoclonal antibody (Acris Antibodies GmbH, Hiddenhausen, Germany) or the pan-HLA Class I monoclonal antibody W6/32 (Dako, Milan, Italy), followed by incubation with a PE-conjugated polyclonal goat anti-mouse antibody (Dako). The fluorescence index (FI) for each peptide was calculated according to the following formula: FI = mean fluorescence intensity (MFI) of T2-A2 cells + peptide/MFI of T2-A2 cells without peptide. The Flu M1 p58–66 peptide was used as positive control and the background expression of HLA-A2 was determined using DMSO as a negative control. Peptides were considered to stabilize HLA-A2 molecules with high affinity when the FI was ≥1.5 and low affinity when the FI was between 1.1 and 1.49. All the assays were repeated a minimum of 3 times, and the results are given as means of replicate experiments.

Expression and purification of recombinant BARF1 protein

The p29 BARF1 protein was expressed in 293-T cells using a recombinant adenovirus system as already described.15 After 24-hr infection, cells were washed 4 times with PBS, then seeded in 4 ml DMEM without serum and phenol, and further incubated for 36 to 48 hr. After harvesting, the cells were centrifuged at 2,000g for 20 min. Recombinant adenovirus and cell debris were eliminated by ultracentrifugation at 16,700g for 2 hr. The culture medium was concentrated using Ultrafree-15 filter device (Millipore). Concanavaline A affinity columns were used to purify BARF1 p29 protein from the concentrated medium. Briefly, the Concanavaline A was diluted in a buffer containing 1 M NaCl, 0.1 M (Na₂HPO₄, NaH₂PO₄) at pH 6, 10⁻³ M (CaCl₂, MgCl₂, MnCl₂) and centrifuged for 15 min at 3,000g. Concentrated medium was then incubated for 18 hr with Concanavaline A. After 4 washes with the same buffer, p29 protein was eluted with 1 M MGP (Methyl α-D-glucopyranoside) for 12 hr.

IFN-γ ELISPOT assay

The IFN-γ release ELISPOT assay was performed using a commercial kit (Human IFN-γ ELISPOT kit, Endogen, Tema Ricerca, Bologna, Italy) according to manufacturer's instructions. The assay was carried out using either 15 μg/ml BARF1 protein-pulsed monocytes as stimulators (5 × 10⁴ cells/well) and purified CD4+ or CD8+ cells as responders (5 × 10⁴ cells /well), or monocytes pulsed with 20 μg/ml peptide and isolated CD8+ T lymphocytes as responders. Autologous monocytes were obtained from PBMC by plastic adherence. For the evaluation of memory responses against BARF1 protein, monocytes were cocultured over-night with recombinant p29 BARF1 protein (15 μg/ml) and the BioPORTER reagent QuickEase™ (Gene Therapy Systems, San Diego, CA). Peptide-loaded monocytes were generated by resuspending the cells in RPMI 1640 10% human AB serum, supplemented with 10 μg/ml of human β₂-microglobulin and aliquoting monocytes into ELISPOT assay plates. Peptides were added at a final concentration of 20 μg/ml and plates were incubated 2 hr at 37°C and 5% CO₂. Purified effectors were obtained by immunomagnetic enrichment protocols using human CD4+ or CD8+ T cell isolation kit II (Miltenyi Biotec, Calderara di Reno, Italy), and were added to protein or peptide-loaded monocytes at the effector:target ratio of 1:1. Monocytes stimulated with tetanus toxoid from clostridium tetani (TT, 5 μg/ml, Calbiochem, San Diego, CA) or FLU M1_58–66 peptide (20 μg/ml) were used as positive controls and and monocytes pulsed with an irrelevant protein (a recombinant VK3-20 immunoglobulin light chain) or peptide served as negative controls. Cells were seeded onto ELISPOT capture plates in quadruplicates and incubated for 48 hr at 37°C and 5% CO₂. All plates were evaluated by a computer-assisted ELISPOT reader (Eli.Expert, A.EL.VIS GmbH, Germany). The number of spots in negative control wells (range: 0–5 spots) was subtracted from the number of spots in stimulated wells. Responses were considered significant if a minimum of 5 IFN-γ producing cells were detected in the well.

Autologous dendritic cell generation and production of peptide-specific CTLs

Purified PBMCs were re-suspended in serum-free medium and incubated at 37°C to allow for plastic-adherent step. After 1 hr, nonadherent peripheral blood lymphocytes were removed and vitally cryopreserved, while immature dendritic cells (DC) were obtained by culture adherent monocytes in complete RPMI 1640 supplemented with 10% of heat-inactivated healthy donors AB serum, recombinant human GM-CSF (50 ng/ml) and IL-4 (25 ng/ml), both from R&D Systems (Abingdon, UK). Cells were substituted with 10% of fresh medium and cytokines concentration were reestablished on Days 3 and 6. On Day 6, DC maturation was also induced with LPS (1.5 μg/ml) from Salmonella Typhimurium (Sigma). After over-night incubation at 37°C, 10⁶ mature DC/ml were seeded in 48-wells plates and pulsed respectively with the p2–10, p23–31 and p49–57 BARF1 peptides (50 μg/ml). About 3μg/ml of human purified β2-microglobulin were added (Chemicon) and then incubated for 2 hr at 37°C in 5% CO₂. Aliquots of peptide-pulsed DC were vitally stored frozen and used in weekly CTL restimulations, while aliquots of 10⁵/ml DC were resuspended in complete medium, supplemented with rhIL-7 (20 ng/ml, R&D Systems) and cocultured in 24-wells plates with autologous peripheral blood lymphocytes, at 10:1 or 20:1 effector:target ratio. rhIL-2 (3 ng/ml, R&D Systems) was added on Day 2, and every 4 days thereafter, to all cultures. After three/four re-stimulation with peptide-pulsed DC, CTLs were tested for cytotoxic activity.

Standard calcein-AM release assay

Cytotoxic activity of peptide-specific CTLs was evaluated using autologous LCLs and the EBV+, BARF1+ Granta 519 cell line as targets in a calcein AM release assay. Target cells (2 × 10⁵) were resuspended in 1 ml of Hanks Balanced Salt Solution without phenol red (HBSS), supplemented with 1% FCS, labelled with 25 μM of calcein-AM (Calbiochem) and incubated 30 min at 37°C 5% CO₂. Labelled cells were washed 3 times and seeded in 96-wells plate at a concentration of 5 × 10³ cells/well. CTLs were added at 20:1, 10:1, 5:1, 2.5:1 and 1.25:1 effector:target ratio. All tests were performed in triplicate. The HLA-A*0201-specific mAb cr11.351 was added to the target cells and incubated at room temperature for 30 min to assess the HLA-A*0201 restriction of CTL responses. To obtain total calcein-releasing cells, targets were incubated with 100 μl/well of lysis buffer (25 mM sodium borate, 0.1% Triton-X100 in HBSS, pH 9.0). Spontaneous release was determined by seeding target cells and adding 100 μl/well of HBSS. Plates were incubated for 4 hr at 37°C and 5% CO₂ in a volume of 200 μl/well. Following incubation, the content of each well was mixed, plates were centrifuged and 100 μl of the supernatant was transferred to a 96-well black culture plate. Fluorescence intensity was measured by reading the plates from the top using a SpectraFluorPlus (Tecan, Austria). Excitation and emission filters were 485 and 535 nm, respectively and gain was set at 70. The percentage of lysis was calculated as follows:

• 

Results

BARF1 mRNA is detected in tumor biopsies but not in normal nasopharyngeal tissues from Italian UNPC patients

As a first step, we aimed at verifying the prevalence of BARF1 expression in a series of UNPC from a nonendemic area such as Italy. Using a specific RT-PCR approach, BARF1 mRNA expression was detected in 15/18 (78%) UNPC biopsies whereas all 8 normal nasopharyngeal tissues analyzed were negative. Notably, 4 of these normal tissues were obtained from UNPC patients whose tumor biopsy was positive for BARF1 mRNA expression. Considering that EBER expression was detected only in UNPC cells and not in infiltrating normal cells, the expression of BARF1 mRNA can be ascribed to tumor cells. Raji cells and the DAA3 LCL were also negative, whereas BARF1 mRNA was constitutively expressed by Akata and Granta 519 cells and upregulated in DAA3 cells after EBV lytic cycle induction (Fig. 1). These findings indicate that BARF1 is expressed by tumor cells of the majority of Italian UNPC.

Figure 1. RT-PCR analysis of BARF1 mRNA gene expression in tumor (T) and normal (N) tissues of UNPC patients. About 80% of UNPC tumor biopsies resulted positive to BARF1 mRNA, whereas all samples of normal nasopharyngeal mucosa analyzed were negative. Raji cells and the DAA3 LCL were also negative, whereas BARF1 mRNA was constitutively expressed by Akata and Granta 519 cells and up-regulated in DAA3 cells after EBV lytic cycle induction.

Detection of BARF1-specific T cell responses in healthy donors and UNPC patients

The presence of spontaneous T cell responses to BARF1 protein was investigated by IFN-γ ELISPOT in PBMCs from 5 healthy donors (3 EBV-seropositive and 2 EBV-seronegative) and 5 UNPC patients (Table I). To this end, autologous monocytes pulsed with recombinant BARF1 p29 protein were incubated for 48 hr with purified CD4+ or CD8+ T cells. As shown in Figure 2, spontaneous CD4+ and CD8+ T cell responses to BARF1 p29 protein were detected in all 3 EBV-seropositive donors investigated (median of 27.2 and 24.3 SFC/50.000, respectively). Conversely, these responses were absent or at the level of those induced by an irrelevant protein (VK3-20 immunoglobulin chain) in EBV-seronegative donors, supporting the EBV specificity of these responses (Fig. 2). Notably, strong CD4+ and CD8+ T cell responses to BARF1 p29 protein were observed in all UNPC cases investigated (Fig. 2). In fact, the extent of these responses was approximately the double of those observed in EBV-seropositive donors, with numbers of responding effectors comparable to those induced by tetanus toxoid (Fig. 2).

Figure 2. Detection of spontaneous CD4+ and CD8+ memory T cell responses to BARF1 protein in EBV-seropositive and EBV-seronegative healthy donors and UNPC patients by enumeration of IFN-γ spot forming cells in ELISPOT assay. Each bar represents the mean ± SD of quadruplicate wells, after subtraction of background responses, given by unpulsed autologous monocytes. Tetanus toxoid and an unrelated protein (VK3-20) were used as positive and negative controls, respectively, and the results are shown as mean ± SD of the SFC/50,000 effectors detected in the donors and UNPC patients analyzed.

Table I. Healthy Donors and UNPC Patients HLA-A and -B High Resolution Genotype

Donors	HLA-A	HLA-B
D1	A*0301/A*6801	B*3501/B*5101
D2	A*0201/A*2901	B*1801/B*4501
D7	A*0101/A*2402	B*0801/B*3502
D12	A*0301/A*3101	B*0702/B*1801
D13	A*0201/A*2402	B*1501/B*3501
UNPC patients
U5	A*0101/A*2901	B*1501/B*3502
U20	A*0301/A*3101	B*5101/B*5701
U21	A*0201/A*0201	B*5001/B*5101
U29	A*3004/A*3201	B*2709/B*5101
U30	A*2402/A*6801	B*0801/B*3801

Selection and binding activity of BARF1 peptides to HLA-A*0201 molecules

Within the first 60 amino acid sequence of the BARF1 protein, immunoinformatic prediction analysis allowed the selection of ten 9-mer peptides that could potentially bind to HLA-A*0201 molecules (Table II). Nevertheless, only 5 of these HLA-A*0201-restricted peptides were available for the study (Table II), since the high hydrophobicity of the p3–11, p6–14, p7–15, p8–16 and p9–17 peptides prevented their successful synthesis and purification. Peptides p2–10, p22–30, p23–31, p29–37 and p49–57 were then tested in the T2 stabilization assay. As shown in Figure 3, the p49–57 peptide demonstrated a low affinity binding in this assay, giving a FI of about 1.2, whereas the other peptides analyzed showed a FI of about 1. In these experiments, the Flu M1_58–66 gave a FI of 2.3, consistently with its high affinity for HLA-A2. The pan HLA Class I monoclonal antibody W6/32 or the anti HLA-A2 BB7.2 antibody gave similar results (not shown).

Figure 3. Peptide-MHC class I stabilization assay using T2-A2 cell line. The fluorescent index (FI) was calculated for each selected peptide, using Flu M1_58–66 peptide as positive control. Peptides were considered to stabilize HLA molecules with high affinity when the FI was ≥1.5 and low affinity when the FI was between 1.1 and 1.49. All the assays were repeated a minimum of 3 times, and the results are given as means of replicate experiments.

Table II. BARF1 Peptides and Predicted Binding to HLA-A*0201

a.a. position	a.a. sequence	SYFPEITHY score	BIMAS score
2–10	RFIAQLLLL	17	0.03
3–11	FIAQLLLLA	21	7.23
6–14	QLLLLASCV	23	257.34
7–15	LLLLASCVA	18	12.81
8–16	LLLASCVAA	21	31.25
9–17	LLASCVAAG	21	0.29
22–30	AFLGERVTL	20	0.05
23–31	FLGERVTLT	21	323.25
29–37	TLTSYWRRV	20	23.65
49–57	KLGPGEEQV	25	119.28

Spontaneous T cell responses against BARF1-derived HLA-A*0201 peptides

Spontaneous CD8+ T cell responses to BARF1-derived peptides (p2–10, p22–30, p23–31, p29–37 and p49–57) were initially evaluated by IFNγ-ELISPOT in 7 HLA-A*0201+ donors (6 EBV-positive and 1 EBV-seronegative), and 3 HLA-A*0201-EBV-seropositive donors. Purified CD8+ T lymphocytes were stimulated with BARF1-derived peptides presented by autologous monocytes. While no BARF1-peptide-specific CD8+ T cell response was observed in HLA-A*0201+ EBV-seronegative or HLA-A*0201-EBV-seropositive donors, all 5 BARF1 peptides elicited variable degrees of specific CD8+ T cell responses in all HLA-A*0201+ EBV-seropositive donors investigated (Table III). Similar results were obtained in the 9 HLA-A*0201+ UNPC patients, although for each peptide the number of IFNγ-secreting CD8+ T cells was significantly higher than detected in EBV-seropositive donors (Tables III and IV). Notably, the immune responses against the p2–10, p23–31 and p49–57 peptides were slightly dominating in both HLA-A*0201+ EBV-seropositive donors and UNPC patients (Tables III and IV).

Table III. Frequency of CD8+ Responders to BARF1-Derived Peptides in Healthy Donors Spots/50.000 CD8+ cells (mean ± SD)

EBV-HLA-A2 donors	p2–10 (RFI)	p22–30 (AFL)	p23–31 (FLG)	p29–37 (TLT)	p49–57 (KLG)	Flu M158–66 (GIL)
D13	3 ± 1	1 ± 1	2 ± 1	0 ± 1	5 ± 1	63 ± 5
EBV+HLA-A2 donors	p2–10 (RFI)	p22–30 (AFL)	p23–31 (FLG)	p29–37 (TLT)	p49–57 (KLG)	Flu M158–66 (GIL)
Purified CD8+ effectors were added to peptide-loaded monocytes at the effector:target ratio of 1:1 and Flu M158–66 peptide was used as positive control. Mean and SD are included.
D1	8 ± 1	5 ± 1	12 ± 1	5 ± 2	13 ± 1	79 ± 11
D2	10 ± 1	4 ± 2	13 ± 1	4 ± 1	12 ± 1	67 ± 7
D3	24 ± 2	15 ± 1	23 ± 2	6 ± 1	23 ± 1	75 ± 12
D4	32 ± 3	21 ± 1	40 ± 3	23 ± 3	44 ± 3	64 ± 9
D5	21 ± 1	23 ± 2	26 ± 1	13 ± 0	21 ± 1	38 ± 3
D6	57 ± 5	31 ± 1	64 ± 4	26 ± 3	74 ± 8	75 ± 7
Average	25	17	30	13	31	66
EBV+HLA-A2^− donors	p2–10 (RFI)	p22–30 (AFL)	p23–31 (FLG)	p29–37 (TLT)	p49–57 (KLG)
D9	0 ± 1	2 ± 2	1 ± 1	1 ± 2	1 ± 2
D10	3 ± 2	1 ± 2	2 ± 2	3 ± 2	1 ± 2
D11	2 ± 2	3 ± 2	2 ± 2	3 ± 2	3 ± 2
Average	2	2	2	2	2

Table IV. Frequency of CD8+ Responders to BARF1-Derived Peptides in UNPC Patients spots/50.000 CD8+ cells (mean ± SD)

UNPC cases	p2–10 (RFI)	p22–30 (AFL)	p23–31 (FLG)	p29–37 (TLT)	p49–57 (KLG)	Flu M158–66 (GIL)
Purified CD8+ effectors were added to peptide-loaded monocytes at the effector:target ratio of 1:1 and Flu M158–66 peptide was used as positive control. Mean and SD are included. Differences between UNPC patients and donors were statistically significant for each BARF1 peptide (p < 0.001).
U 16	93 ± 12	15 ± 2	109 ± 17	76 ± 8	97 ± 12	127 ± 14
U 21	77 ± 11	41 ± 9	95 ± 6	37 ± 3	103 ± 14	121 ± 11
U 22	159 ± 12	83 ± 11	139 ± 12	95 ± 6	143 ± 9	195 ± 15
U 23	61 ± 6	21 ± 2	68 ± 6	12 ± 2	65 ± 6	73 ± 6
U 24	86 ± 4	104 ± 13	107 ± 16	76 ± 5	101 ± 8	153 ± 11
U 25	160 ± 9	157 ± 12	166 ± 9	148 ± 13	200 ± 6	183 ± 17
U 26	23 ± 2	17 ± 2	26 ± 2	9 ± 3	25 ± 3	40 ± 5
U 27	32 ± 3	14 ± 3	36 ± 2	12 ± 3	44 ± 3	49 ± 4
U 28	27 ± 2	13 ± 1	32 ± 7	9 ± 1	34 ± 3	32 ± 2
Average	72	47	78	46	81	97

Induction of peptide-specific CTLs and cytotoxic activity of CTL cultures

Peptide-specific CTLs were generated from 2 HLA-A*0201, EBV-seropositive healthy donors by stimulating peripheral blood T lymphocytes with autologous DC loaded with the p2–10, p23–31, or p49–57 peptides. After 3 weekly restimulations, cytotoxic activity of CTL cultures was evaluated using autologous peptide-loaded LCLs as targets in calcein-AM cytotoxicity assays. As shown in Figure 4a, CTL cultures showed HLA-A*0201 peptide-specific killing: T cells only recognized autologous LCLs loaded with the respective BARF1 HLA-A*0201 peptide, whereas no lysis was induced in unpulsed LCLs. Moreover, preincubation with an anti-HLA-A2 antibody markedly inhibited the specific lysis, indicating that the cytotoxic activity of these effectors was HLA-A2-restricted. In the next set of experiments, we evaluated the ability of BARF1 peptide-specific CTLs to lyse tumor cells endogenously expressing the BARF1 protein. To this end, we used as a target the HLA-A*0201+, EBV-carrying Granta 519 cell line constitutively expressing BARF1. CTLs obtained from two different donors (sharing only the HLA-A*0201 allele with Granta 519 cells) and specific for the p2–10, p23–31 or p49–57 BARF1 peptides were able to efficiently lyse Granta 519 cells in an HLA-A2-restricted fashion, as shown by inhibition of cytotoxicity in the presence of a monoclonal antibody to HLA-A2 (Fig. 4b). The extent of specific killing was similar to that induced in peptide-pulsed autologous LCLs (Fig. 4b). Again, no lysis of autologous unpulsed LCLs was observed, confirming the specificity of killing (Figs. 4a and 4b). It is worth mentioning that all the LCLs used in these experiments as antigen presenting cells were basally negative for BARF1 mRNA expression, consistently with the notion that less than 1–2% of cells in these cultures spontaneously undergo lytic cycle induction, being thus able to express BARF1. The number of BARF1+ cells presumably present in this culture is probably too low to affect the results of these cytotoxicity tests. These data indicate that DC pulsed with BARF1-derived peptides can induce BARF1-specific Class I-restricted CTLs that recognize tumor cells endogenously expressing BARF1 in an antigen- specific and HLA-A-restricted manner in vitro. These findings also indicate that the p2–10, p23–31 and p49–57 BARF1 peptides can be processed naturally in cancer cells and can be efficiently presented in the context of HLA-A*0201 on the cell surface of cancer cells to be recognized by CTLs.

Figure 4. CTL cultures from 2 HLA A*0201+ healthy donors were generated against 3 BARF1-selected peptides. (a) Using peptide-loaded and un-pulsed autologous LCLs as target cells, specific cytotoxic activity was evaluated by standard calcein-AM release assay. All tests were performed in triplicate and at Effector:Target Ratio of 20:1, 10:1, 5:1, 2.5:1 and 1.25:1. (b) Peptide-specific cytotoxic T cells were also able to efficiently lyse tumor cells endogenously expressing the BARF1 protein, in an HLA-A2-restricted fashion, as demonstrated by the incubation of target cells with the anti-HLA-A*0201 cr11.351 mAb. No lysis of autologous un-pulsed LCLs was observed in all performed tests, thus confirming the specificity of killing.

Discussion

Although integrated treatment with radiotherapy and chemotherapy has considerably improved the clinical control of UNPC, a significant number of patients relapse, particularly those with advanced disease at diagnosis.21 Moreover, conventional treatments are often accompanied by acute and late severe side effects that may compromise patient quality of life. Second primary cancers, including EBV-associated tumors, are an emerging cause of death in large cohort series.21 Therefore, there is a pressing need to develop new and less toxic therapies able to improve survival of UNPC patients. Immunotherapy approaches targeting EBV-encoded viral antigens appear promising in this respect, having shown the ability to induce specific immune responses and objective clinical responses, although only in a fraction of UNPC cases.6, 8

The BARF1 protein may constitute a potentially useful target to improve the efficacy of EBV-based immunotherapy for UNPC, since this viral antigen is exclusively expressed in EBV-associated carcinomas.9, 10, 22 In the present study, we have verified that BARF1 is expressed also in UNPC from a nonendemic region such as Italy. In fact, 80% of UNPC from our series selectively expressed BARF1 mRNA in tumor biopsies, whereas normal nasopharyngeal tissues were negative. Since EBER expression was detected only in UNPC cells and not in infiltrating normal cells, the expression of BARF1 mRNA can be bona fide ascribed to tumor cells. We also provide evidence indicating that BARF1 is naturally immunogenic for T lymphocytes, as shown by the detection of spontaneous CD4+ and CD8+ T cell responses specific for the p29 BARF1 protein in both EBV-seropositive donors and UNPC patients.

Using immunoinformatic prediction tools, we have selected and characterized 5 different candidate BARF1 CTL epitopes presented by HLA-A*0201, the most frequent HLA-A allele in Caucasians. All these 5 epitopes mapped within the first 60 amino acids of the BARF1 protein; a region that includes the conserved transforming domain. Using the T2-A2 stabilization assay, however, only 1 (p49–57) of the 5 peptides analyzed was able to bind HLA-A2, although with low affinity. Despite these premises, all 5 BARF1 nonamers available for the study readily elicited specific CD8+ T cell responses in EBV-seropositive HLA-A*0201+ donors, as shown by IFNγ-ELISPOT. In all donors investigated, immune responses against the p2–10, p23–31 and p49–57 peptides tended to be slightly dominating, with a mean frequency of peptide-specific CD8+ T cells corresponding to about half of that observed for the control FLU M1_58–66 peptide. These findings are consistent with the notion that higher peptide binding affinity for HLA molecules does not necessarily correspond to the functional activity of the responding T lymphocytes.23, 24 The direct detection of strong BARF1-specific responses without prestimulation in EBV-seropositive donors suggests the existence of a relatively high frequency of BARF1-specific CD8+ T cell precursors, consistently with what observed for other EBV-encoded proteins, particularly those expressed during EBV lytic replication.25, 26 Considering that in normal EBV-infected B cells BARF1 is expressed as an early gene,27 the high frequency of BARF1-specific T lymphocytes detected in healthy donors is consistent with a likely relevant role of these effectors in keeping a tight control of EBV replication.

Notably, the magnitude of CD8+ T cell responses to the whole BARF1 protein and derived peptides was significantly higher in UNPC patients than in healthy donors. This contrasts with what observed for the latent EBV proteins LMP-1 and LMP-2 whose subdominant epitopes elicited weaker responses in UNPC patients as a possible consequence of tumor-related immune suppression.5, 28 A possible explanation for this apparent discrepancy may be found in the biological properties of the BARF1 protein, which, unlike LMP-1 and LMP-2, can be actively secreted by UNPC cells. Within tumor microenvironment, therefore, the BARF1 protein may be captured and processed by local dendritic cells or other antigen presenting cells, which may provide a sustained presentation of antigenic BARF1 peptides. This may lead to the increased numbers of BARF1-specific T cells circulating in the blood of UNPC patients, although it remains to be elucidated whether these effectors are detectable and functional within UNPC lesions.

We have also verified whether the identified BARF1 peptides could be used as T cell epitopes able to induce antigen-specific CTLs. This has been accomplished using PBMCs from HLA-A*0201+ donors in an in vitro immunization protocol including autologous monocyte-derived DC pulsed with the BARF1 peptides as antigen presenting cells. Our results show that CTLs specific for the p2–10, p23–31, or p49–57 BARF1 peptides can be easily induced from PBMCs of EBV-seropositive healthy donors. These cultures were not only able to lyse autologous LCLs loaded with the antigenic peptide, but also recognized tumor cells endogenously expressing BARF1 in an antigen-specific and HLA-A2-restricted manner. These findings indicate that the p2–10, p23–31 or p49–57 BARF1 peptides used for CTL induction are also naturally processed and presented as CTL epitopes by BARF1-expressing tumor cells.

The identification and validation of HLA-A*0201-restricted BARF1 epitopes will allow the synthesis of tetramers potentially useful to enumerate and characterize the differentiation features of epitope-specific T lymphocytes in UNPC patients. This may be relevant not only to disclose possible prognostic correlates in this setting, but also to monitor BARF1-specific immune responses in patients treated with EBV-targeted immunotherapy protocols.

In conclusion, our results demonstrate that BARF1 provides target epitopes for spontaneous T cell responses in EBV-seropositive healthy donors and, to a greater extent, in UNPC patients. These findings, together with the previous observation that BARF1 also elicits strong humoral responses in UNPC patients,18 indicate that BARF1 is a particularly attractive antigen with immunogenic properties in most UNPC patients. This provides the rational background to verify whether boosting immune responses also towards BARF1 may improve the extent and duration of clinical responses induced in UNPC patients by immunotherapy protocols.

Acknowledgements

We are grateful to Dr. Andrea Anichini, Milan, for providing us with the T2-A2 cell line and the cr11.351 antibody. We also thank the donors and patients who participated in this study.