Defining the expression hierarchy of latent T-cell epitopes in Epstein-Barr virus infection with TCR-like antibodies

Epstein-Barr virus (EBV) is a gamma herpesvirus that causes a life-long latent infection in human hosts. The latent gene products LMP1, LMP2A and EBNA1 are expressed by EBV-associated tumors and peptide epitopes derived from these can be targeted by CD8 Cytotoxic T-Lymphocyte (CTL) lines. Whilst CTL-based methodologies can be utilized to infer the presence of specific latent epitopes, they do not allow a direct visualization or quantitation of these epitopes. Here, we describe the characterization of three TCR-like monoclonal antibodies (mAbs) targeting the latent epitopes LMP1125–133, LMP2A426–434 or EBNA1562–570 in association with HLA-A0201. These are employed to map the expression hierarchy of endogenously generated EBV epitopes. The dominance of EBNA1562–570 in association with HLA-A0201 was consistently observed in cell lines and EBV-associated tumor biopsies. These data highlight the discordance between MHC-epitope density and frequencies of associated CTL with implications for cell-based immunotherapies and/or vaccines for EBV-associated disease.

including flow cytometry, immunohistochemical staining, and confocal microscopy. We next established the epitope expression hierarchy amongst the three latent epitopes in cell lines and clinically relevant EBV-associated tumor biopsies. Our observations of this hierarchy and its differential binding on strain-associated epitope variants have important implications for diagnosis, immuno-targeting and vaccine development.

Results
Generation of high affinity TCR-like mAbs with exquisite specificity. In this report, we highlight an adaptation of conventional hybridoma technology that enabled the production of high-affinity TCR-like mAbs targeting three EBV latent epitopes displayed on HLA-A0201. The methodology is illustrated in Supplementary  Information (Supplementary Fig. S1). Briefly, membrane-free HLA-A0201 associated with EBV latent peptides (EBNA1 562-570 : FMVFLQTHI; LMP1 125-133 : YLLEMLWRL; LMP2A 426-434 : CLGG-LLTMV) were generated to immunize mice using an established protocol [8][9][10][11] . The splenocytes of immunized mice were immunomagnetically selected prior to fusion. It is only with this enrichment that hybridomas producing TCR-like mAbs targeting EBNA1 562-570 and LMP2A 426-434 in association with HLA-A0201 can be generated (Fig. 1a). For hybridomas producing antibodies targeting LMP1 125-133 in association with HLA-A0201, there is an increase in the percentage of such hybridomas isolated following splenocytes enrichment. The optimal representative monoclonal hybridoma for each target was selected for subsequent analyses.
The binding specificities of these antibodies were examined using flow cytometric analysis of T2 cells pulsed with 12 different HLA-A0201 restricted peptides. The mAbs exhibited exquisite specificity for their respective target peptide and not other HLA-A0201 restricted epitopes from a variety of human pathogens (Fig. 1b).
To further examine the specificity of each TCR-like mAb in comparison to specific CTL, we determined their ability to inhibit CTL lysis. The three mAbs inhibited the activity of their respective CTLs in a specific dose-dependent manner, as shown by the inhibition of CTL-inflected 51 Cr release from target cells (Fig. 1d). Thus, we can infer a degree of overlap in the targeting of TCRs and TCR-like mAbs for the same viral epitopes. With these mAbs endowed with TCR specificity, we can visualize and quantitate the expression profile of latent EBV epitopes in infected cells.
Epitope variants are differentially recognized by respective TCRlike mAbs. A factor that impacts upon epitope presentation is strain differences in the encoding sequence of the latent antigens that translates to CTL epitope variants. The classical methodology of EBV typing does not adequately distinguish the pathogenic/tumorigenic nature of various virus strains 12 . This typing does not give due consideration to the clinical diversity of EBNA1, LMP1, and LMP2A, which are the only three latent genes observed in associated malignancies 1 . Using alternative classifications that were adopted for each of these genes, we investigated the presentation of clinical epitope variants [9][10][11] . Each of these epitope variants was pulsed onto HLA-A0201 homozygous B cell lymphoblastoid cell lines for flow cytometric analysis using the three TCR-like mAbs (Fig. 1e). Anti-HLA-A02/EBNA1 562-570 antibody can bind the B95-8 and Raji variants. Anti-HLA-A02/LMP1 125-133 antibody exhibited good binding activity for B95-8 and Mediterranean (Med) variants but none was observed for cells pulsed with the China1 variant-the form of EBV that is linked to NPC 10 . Strong binding was observed for anti-HLA-A02/LMP2A 426-434 antibody for both the B95-8 and the NPC variants. With the exception of the LMP1 (China1) variant, our TCR-like mAbs can recognize the clinical variants of our epitopes of interest (Fig. 1e). Together, these findings indicate that epitopes associated with HLA-A0201 on infected cells are differentially presented and this can be influenced by mutations associated with strain variation.
Mapping the expression hierarchy of three latent epitopes on EBV infected cells. We next analyzed the endogenous expression of EBNA1 562-570 , LMP1 125-133 and LMP2A 426-434 epitopes on three HLA-A0201 EBV-infected tumor cell lines of lymphoid and epithelial origins: CCRF-SB, RPMI-6666 and C666-1A2 13 . The epitope variant of each cell line was determined by sequencing their associated EBV virus (Fig. 2a). A flow cytometry based beadcalibration kit, QIFIKITH (Dako), allowed the quantitation of the three EBV latent peptides presented by HLA-A0201 (Supplementary Fig. S2). This enabled us to map the ligand density and expression hierarchy amongst the three latent EBV epitopes in the cell lines (Fig. 2b). HLA-A0201/EBNA1 562-570 complexes were strongly expressed by the RPMI-6666 and CCRF-SB cell lines This represents the first analysis of an expression hierarchy of HLA-A0201/peptide complexes of latent EBV antigens. Surprisingly, the high expression and dominance of HLA-A0201/EBNA1 562-570 complexes was consistent for all cell lines tested. With the exception for the LMP1 (China1) variant of C666-1A2, our mAbs could bind the variants of each epitope. This was consistent with our previous data on binding of epitope variants (Fig. 1e).
The mAbs were also utilized to determine the cellular localization and relative expression of the 3 latent EBV HLA-A0201/peptide complexes by confocal microscopy. A high degree of surface localization of the HLA-A0201/EBNA1 562-570 complexes was observed in all three cell lines ( Supplementary Fig. S4) and the staining intensities of the three latent antigens correlated with the quantitation data shown in Figure 2b. In all cases, expression of HLA-A0201/EBNA-1 562-570 complexes on the surface of the tumor cells was stronger than/LMP-1 125-133 or/LMP-2A 426-434 . Our observations cannot be ascribed to the presence or absence of the latent proteins from which the epitopes were derived, as indicated by RT-PCR and immunoblot ( Figure 2c). The absence of an LMP1 immunoblot for C666-1A2 was due to the mutations in the LMP1 (China1) variant that altered the binding site for the anti-LMP1 antibody used 13 .
Detection of latent viral epitope expression in a humanized EBV mouse model. To analyze the expression of the three antigens in the context of a natural EBV infection, we employed an EBV-infected humanized NOD/SCID/IL2rc null (Hu-NSG) mouse model based on reported protocols 14 . The latency program exhibited by these mice was type III and thus the same as that observed in lymphoproliferative disorders in immuno-compromised patients 1 .
A schematic of our experimental approach is provided (Fig. 3a). Following hematopoietic stem cells (HSC) reconstitution in NSG mice and infection with EBV, the spleens were harvested for staining. Immunofluorescence staining was carried out using R-Phycoerythrin (PE) conjugated TCR-like mAbs (Fig. 3b). Due to the haplotype specific nature of our mAbs, only mice engrafted with HLA-A0201 HSC exhibited staining for all three complexes (Fig. 3b). Thus, these staining validated the employment of our TCR-like mAbs for epitope presentation studies in EBV-infected Hu-NSG mouse model. Dominance of EBNA1 562-570 epitope expression in EBV-tumor biopsies. To determine the clinical relevance of our observations, cryo-sections of HLA-A0201 EBV-positive NPC biopsies were stained with the respective TCR-like mAbs. EBV infected cells in these biopsies stained strongly for HLA-A0201/EBNA1 562-570 complexes and weakly for HLA-A0201/LMP2A 426-434 complexes. Sections were also stained with BB7.2 to determine their expression of HLA-A02 molecules. A representative data set (from 5 patients) is shown (Fig. 3c). The staining intensity correlated with that observed in the NPC cell line C666-1A2 ( Supplementary Fig.  S3). The LMP1 125-133 complex staining was not observed due to epitope mutation in the EBV NPC-linked China1 strain, correlating with the data in Figure 1e.

Discussion
TCR-like mAbs are valuable reagents for direct epitope visualization and detection. Several studies had illustrated its usefulness in epitope discovery 15 , epitope processing 16 , and immuno-targeting 17 . However, the numbers of documented antibodies are limited.
Previous reports documenting the generation of these mAbs used either classical hybridoma fusion following murine immunization or phage display methodologies. These resulted in antibodies that are relatively low in affinity 18 and stain poorly for endogenously generated epitopes 19 .
By enriching for immunized splenocytes prior to fusion, we increased the number of hybridomas producing antibodies targeting LMP1 compared to non-enrichment. Moreover, it is only with this  enrichment approach that hybridomas producing two of our TCRlike mAbs targeting EBNA1 and LMP2A could be isolated. Futhermore, our antibodies exhibit high specificity for their respective epitopes and binding affinity in the nanomolar range. Our data validates this approach for isolating such rare hybridomas 5 that are technically challenging to generate 20,21 . The high affinity and exquisite specificity of these reagents enabled us to examine epitope variants presented on HLA-A0201 BLCLs. These epitope mutations are due to viral sequence variation observed in each of the latent EBV genes (LMP1, LMP2A and EBNA1). Our antibodies bind to the epitope variants with differential activity suggesting that the associated amino acid substitutions impact upon presentation. Given that B95-8 is the only GMP-grade EBV strain employed for the expansion of specific CTL for adoptive immunotherapy, our data indicate that more attention must be paid to the EBV strain infecting the patient to ensure in vivo efficacy of the adoptively transferred T-cells.
Staining for endogenously produced epitopes/peptides is a major challenge for TCR-like mAbs 19 . Therefore, it was necessary to investigate the binding ability of our mAbs for its respective endogenously generated cognate peptide/MHC. As observed, our reagents enabled the visualization of intracellular and extracellular localization of endogenously generated epitopes. These reagents were also used to visualize latent EBV epitope expressing cells in an EBV-infected humanized mouse model. Coupled with QIFIKITH, we also quantitated the density of endogenously generated epitopes in these three EBV infected cell lines. A distinct epitope expression hierarchy was consistently observed for all three cell lines with the dominance of EBNA1 562-570 over the other two epitopes analyzed. This was similarly observed in clinically relevant EBV-associated NPC biopsies. The dominance of EBNA1 562-570 epitope did not correlate with the reported frequency of the respective specific CTLs 22 where CTLs targeting LMP2A 426-434 are dominant over the other two epitopes (EBNA1 562-570 and LMP1 125-133 ).
Previous reports have intimated that EBNA1 inhibits its own processing and presentation 23 . However, epitope expression can be inferred from the presence of EBNA1 specific CTLs-this has been proposed as evidence for dendritic cells cross-presentation of EBNA1 peptides 24 . Despite such indications, we are the first to directly visualize and detect HLA-A0201/EBNA1 562-570 complex expression in infected cells of both hematopoietic and non-hematopoietic origin. This suggests that EBNA1 can be processed and presented through the classical MHC class I pathway. Although the presentation of HLA-A0201/EBNA-1 562-570 was reported to be low or non-existent in previous reports employing CTL lines 21,25,26 , we were able to directly detect this ''rare'' epitope in relatively high numbers. This discordance could be due to dysfunctional CTLs 4 or suggest that the density of peptide-MHC complexes is not necessarily the key factor in CTL selection and/or dominance. The immunodominance of CTLs can be affected by an interplay of cellular and molecular factors such as CTL precursor frequencies 27 , quality of TCR-peptide-MHC signaling potency 28 and the relative binding affinity of the peptide epitope for its associated MHC molecule 29 . These components and in vivo tumor microenvironment 4 can influence both the scale and form of the CTL response but were not addressed as part of this study.
In summary, by employing TCR-like mAbs, we have determined the expression hierarchy of EBV latent antigens. Our data indicates that EBNA1 562-570 and LMP2A 426-434 are possible biomarkers of EBV-associated tumors in HLA-A0201 patients. The dominance of EBNA1 562-570 amongst the three epitopes was validated in cell lines and ex vivo staining of clinical NPC biopsies. This observation has important implications for both immunotherapy and EBV vaccine development. Influenza A virus M1 58-66 (GILGFVFTL) were synthesized by Mimotopes to 95% purity and was analyzed by electrospray mass spectrometry. The murine IgG1 isotype control antibody MOPC 21 was purchased from Sigma-Aldrich. The rabbit anti-beta-2 microglobulin polyclonal antibody and horseradish-peroxidase-conjugated goat anti-mouse IgG were purchased from Thermo Fisher Scientific. CS1-4 antibody cocktail (anti-EBV LMP1 antibody) and PE-conjugated goat anti-mouse IgG (H 1 L) were purchased from Dako. Alexa FluorH 488-conjugated goat anti-mouse IgG (H 1 L) antibodies were purchased from Invitrogen. 14B7 (anti-EBV LMP2 antibody) and E1-2.5 (anti-EBV EBNA1 antibody) were purchased from Abd Serotec. Anti-human CD20 antibody was purchased from Abcam.

Methods
Surface plasmon resonance (SPR). Kinetic studies were evaluated by surface plasmon resonance using a BIAcore 3000 TM (GE Healthcare). The monoclonal antibodies were immobilized covalently onto the surface of sensor chip CM5 by amine coupling. To determine the dissociation constant (K D ) of the antibodies, the respective MHC class I/peptide were flowed at different concentrations ranging from 320 nM to 10 nM over the relevant wells at a rate of 30 mL min 21 at 25uC. Responses were recorded and analyzed using BIAevaluation software 3.2. Western blot. CCRF-SB, RPMI666, and C666-1A2 cells were cultured in 100 mm dishes and lysed with 1 3 RIPA buffer (Thermo Fisher Scientific) supplemented with protease inhibitors (Roche Applied Science). The lysates were boiled for 15 min and separated by 10% SDS-polyacrylamide gel electrophoresis, and transferred onto polyvinylidene difluoride membrane (GE Healthcare) overnight. The membranes were probed with CS1-4 antibody cocktail (anti-EBV LMP1 antibody) (152000 dilution) (Dako), 14B7 (anti-EBV LMP2A antibody) (155000 dilution) (AbD Serotec) or E1-2.5 (anti-EBV EBNA1 antibody) (155000 dilution) (AbD Serotec). The membranes were subsequently incubated with horseradish-peroxidaseconjugated goat anti-mouse IgG or goat anti-rat IgG (Thermo Fisher Scientific). The proteins were detected with enhanced chemiluminescence reagent (Western Lightning ECL kit, Perkin Elmer).
Detection of LMP1, LMP2A and EBNA1 transcripts by RT-PCR. RNA from 1 3 10 6 cells was extracted using High Pure RNA Isolation kit (Roche Diagnostics) and 1 mg of total RNA was amplified by RT-PCR with OneStep RT-PCR kit (Qiagen) using the appropriate primers as previously reported 30 . Beta actin was used as an internal control.
Quantitation assays. Epstein-Barr virus infected HLA-A02 cells were stained with the various TCR-like mAb followed by PE-conjugated goat-anti-mouse IgG (Dako). The level of fluorescence intensity was compared with that of calibrated beads containing pre-determined quantity of mouse monoclonal antibody per bead (QIFIKITH calibration beads; Dako). By means of a standard curve based upon the calibration beads, the number of complexes on the cell surface was determined from the fluorescence intensity of each TCR-like monoclonal antibody.
Raising of specific CTL lines. The HLA-A0201 restricted CTL specific for peptides of Epstein-Barr associated proteins -LMP1 125-133 (YLLEMLWRL), LMP2A 426-434 (CLGGLLTMV) or EBNA1 562-570 (FMVFLQTHI) was prepared. CD81 T cells were isolated from a healthy HLA-A0201 positive donor. Human peripheral blood mononuclear cells (PBMC) were isolated from fresh heparinized blood from healthy donor by Ficoll-Paque density gradient centrifugation and CD81 T cells were subsequently isolated using MACs isolation of CD81 positive T cells (Miltenyi Biotec) in AIM H V media supplemented with 2% human AB serum. These isolated CD8 T cells were specifically expanded by adding it into wells seeded with 1 mM of LMP1 125-133 (YLLEMLWRL), LMP2A 426-434 (CLGGLLTMV) or EBNA1 562-570 (FMVFLQTHI) peptides pulsed irradiated PBMC feeder cells and incubated at 37uC in a CO 2 incubator. On day 2 after isolation, the media of the wells were supplemented with 20 U/ml IL2 to expand specific peptide stimulated CD81 T cells. On day 10, CTLs were cloned by limiting dilution assay and expanded in AIM H V media supplemented with 2% AB serum, 20 U/ml IL2, 5 ng/ml IL7 and 100 pg/ml IL12 (R&D Systems) at a density of 500,000 cells/well in 96-well plates seeded with irradiated PBMC feeder cells pulsed with 1 mM of respective specific peptides.
Confocal microscopy. RPMI-6666 and CCRF-SB were seeded onto poly-L-lysine coated glass coverslips while C666-1A2 was seeded overnight on glass coverslips before fixing with methanol-acetone solution. The fixative was quenched with PBS-10 mM glycine before staining was carried out using the respective TCR-like mAbs or rabbit anti-beta-2-microglobulin polyclonal antibody (Thermo Fisher Scientific). Staining was performed in a humidity chamber for 1 h in permeabilization buffer with the concentrations of antibody at 151000 dilutions. The coverslips were subsequently stained with either PE-conjugated goat antimouse IgG (H 1 L) (Dako) or Alexa FluorH 647-conjugated goat anti-rabbit IgG (H 1 L) (Invitrogen) with DAPI. ProLong Gold Antifade reagent (Invitrogen) was used to mount the slides prior to analysis with a Leica TCS SP5 confocal microsystems (Leica Microsystems) using the HCX PLAPO 63 3 objective (numerical aperture: 1.4) The images were acquired at a resolution of 1024 3 1024 pixels for each set. Chromium-release cytotoxic assay. 3 3 10 5 C1R-A2 cells were incubated with 50 mCi 51 Cr (Perkin Elmer) in 100 ml culture medium, with or without the relevant peptide (5 mM) for 1 h at 37uC, and plated in 96-well culture plates (3,000 cells per well). HLA-A02 restricted epitope specific CTL lines were then incubated for 3.5 h in 200 ml culture medium at an E/T ratio of 1051. Supernatant was harvested from each well and transferred onto a lumaplate (Perkin Elmer) for analysis with a TopCount NXT (Perkin Elmer). The percentage of release was calculated using the formula: average of sample release~spontaneous release=total release À spontaneous release ð Þ|100 TCR-like mAb staining of Hu-NSG mouse tissues. Spleens were harvested from EBV-infected Hu-NSG mice and frozen in OCT medium. Tissue cryosections were subsequently prepared on poly-L-lysine coated slides. The sections were fixed and incubated with anti-human and anti-mouse FcR blocking agents (Miltenyi Biotec) to prevent non-specific binding of the antibodies. The tissues were stained with PEconjugated TCR-like mAbs for 2 h. After washing, each slide was coated with DAPI containing SlowFade Gold solution (Invitrogen). The slides were scanned and viewed with MiraxViewer 1.12 software (Carl Zeiss).
Immunohistochemistry. Frozen NPC biopsies were sectioned (4 mm) and embedded onto slides. Slides were fixed and subsequently stained with TCR-like mAbs or BB7.2 31 . Peroxidase block (Dako) was used prior to treatment with Dako labeled polymer. After further washing, immunoperoxidase staining was developed with DAB (Dako).
Animal experiments. BALB/c and NOD/SCID/IL2rc null (NSG) mice were housed in appropriate facilities. All experiments were approved by the IACUC of the National University of Singapore and conducted in accordance with the Guide for the Care and Use of Laboratory Animals (IACUC no: 027/07 and 106/10).
Biopsy and blood collection. Informed consent from healthy volunteers and patients were taken before biopsy or blood extraction. NPC patients at the National University Hospital were enrolled with informed consent. The Institutional Review Board of the hospital approved this study (IRB no: 07-043E).
Statistical analysis. Empirical data were analyzed using GraphPad Prism, version 5.0. 1-way ANOVA with Bonferroni post-test.