Identification of Human T Cell Receptor γδ-recognized Epitopes/Proteins via CDR3δ Peptide-based Immunobiochemical Strategy*

Human T lymphocytes, bearing T cell receptor (TCR) γδ, play an important role in anti-tumor/microbe immune responses. However, few tumor antigens recognized by TCRγδ have been defined so far. To investigate antigenic epitopes/proteins recognized by γδT cells, we have established a new immunobiochemical strategy that uses complementarity-determining region 3 of TCR δ chain (CDR3δ) peptide-mediated epitope/protein-binding assays. CDR3δ peptides synthesized using the CDR3 region in TCR Vδ2 chain were validated for their binding specificity to target cells or tissues. These CDR3δ peptides were then employed as probes to pan putative epitopes in a 12-mer random peptide phage-displayed library and to identify putative protein ligands within tumor protein extracts by affinity chromatography and liquid chromatography/electrospray ionization-tandem mass spectrometry analysis. As a result, we have identified nine peptides and two proteins for TCRγδ, including human mutS homolog 2 (hMSH2) and heat shock protein (HSP) 60. All nine tested epitope peptides not only bind to γδT cells but also functionally activate γδT cells in vitro. Identification of HSP60 confirms the validity of this method as HSP60 is an identified ligand for TCRγδ. We show that hMSH2 is expressed on the surface of SKOV3 tumor cells, and cytotoxicity of Vδ2 γδT cells to SKOV3 cells was blocked by anti-hMSH2 antibody, suggesting that hMSH2 may be a new ligand for TCRγδ. Taken together, our findings provide a novel immunobiochemical strategy to identify epitopes/proteins recognized by γδT cells.

Human T lymphocytes, bearing T cell receptor (TCR) ␥␦, play an important role in anti-tumor/microbe immune responses. However, few tumor antigens recognized by TCR␥␦ have been defined so far. To investigate antigenic epitopes/proteins recognized by ␥␦T cells, we have established a new immunobiochemical strategy that uses complementarity-determining region 3 of TCR ␦ chain (CDR3␦) peptide-mediated epitope/protein-binding assays. CDR3␦ peptides synthesized using the CDR3 region in TCR V␦2 chain were validated for their binding specificity to target cells or tissues. These CDR3␦ peptides were then employed as probes to pan putative epitopes in a 12-mer random peptide phage-displayed library and to identify putative protein ligands within tumor protein extracts by affinity chromatography and liquid chromatography/electrospray ionizationtandem mass spectrometry analysis. As a result, we have identified nine peptides and two proteins for TCR␥␦, including human mutS homolog 2 (hMSH2) and heat shock protein (HSP) 60. All nine tested epitope peptides not only bind to ␥␦T cells but also functionally activate ␥␦T cells in vitro. Identification of HSP60 confirms the validity of this method as HSP60 is an identified ligand for TCR␥␦. We show that hMSH2 is expressed on the surface of SKOV3 tumor cells, and cytotoxicity of V␦2 ␥␦T cells to SKOV3 cells was blocked by anti-hMSH2 antibody, suggesting that hMSH2 may be a new ligand for TCR␥␦. Taken together, our findings provide a novel immunobiochemical strategy to identify epitopes/proteins recognized by ␥␦T cells.
T lymphocytes are classified structurally into two types according to different T cell receptors (TCR) 3 ␣␤ or ␥␦. TCR␣␤ specifically recognize antigenic peptides (epitopes) presented by MHC I/II molecules. Although many peptide ligands have been identified for TCR␣␤ (1), only a few ligands have been identified for TCR␥␦ so far (2,3).
Human ␥␦T cells account for ϳ5% of CD3 ϩ T cells in the peripheral blood, but constitute a major T cell subset in other anatomic locations, such as the intestine. In the peripheral blood of healthy individuals, TCR V␦2 chain pairing with one particular V␥9 chain is expressed on 50 -90% of the ␥␦T cells, whereas intestinal intraepithelial ␥␦T cells frequently express the V␦1 gene, which can associate with different V␥ elements (4). The antigen-binding site of TCR␥␦ is formed primarily from three complementarity-determining regions (CDRs) contributed by each V␥ and V␦ domain. CDR1 and CDR2 fragments are encoded by germ line V genes, whereas the CDR3 is formed by somatic rearrangement of V(D) and J fragments. Sequence diversity in antigen receptors is not evenly distributed among all six CDRs but is highly concentrated in one or two CDR3. It had been proposed that the principal antigen specificity of an immunoglobulin or TCR is derived from its most diverse CDR3 fragments (5).
In addition to TCR diversity, structures of the TCR-CD3 complex, and tissue distribution, the pattern of antigenic recognition is another key difference between ␥␦T cells and conventional ␣␤T cells. TCR␥␦T cells have two important features in their antigenic recognition. First, TCR␥␦ is structurally similar to the B cell receptor (BCR) on B lymphocytes, binding to its antigens or epitopes directly to trigger a fast response against foreign pathogens or self-malignant cells (6 and Toll-like receptors (TLRs) are functionally alike, because these receptors recognize conserved molecules (7). Although ␥␦T cells only account for a small proportion in the human T lymphocyte pool, they play important roles in both anti-tumor and anti-microbe responses in innate immunity (8). Human ␥␦T cells recognize stress-induced or malignancy/infection-related antigens. Such recognition is mediated through a non-MHC restricted manner in most cases. Among the known antigens for ␥␦T cells, the most frequently identified ones are non-peptide antigens, such as pyrophosphoantigen (naturally occurring and synthetic antigens) recognized by T cells bearing V␥9␦2 heterodimer TCR chain (9 -11). In addition, classical MHC and MHC-like molecules can directly serve as ligands for ␥␦T cells (12). MHC class I chain-related gene A (MICA) and B (MICB) are recognized by V␦1 ␥␦T cells (13,14). UL-16 binding protein (ULBP) is a novel ligand for NKG2D receptor in humans (15). We found that RAET1E2, a soluble isoform of the UL16-binding protein RAET1E produced by tumor cells, inhibits NKG2D-mediated NK cytotoxicity (16). Because of the sequence and structure similarity of the extracellular domains between MICA and ULBPs, ULBPs are also considered as antigens for V␦1 ␥␦T cells (17). Heat shock proteins (HSP) are highly conserved among prokaryotes and eukaryotes and are known to be involved in various stress conditions. HSP had been implicated in ␥␦T cell-mediated antitumor response (18,19). Moreover, other protein antigens, such as ectopically expressed mitochondrial ATPase, have also been identified as ligands for TCR␥␦ (20).
However, a crucial issue regarding ␥␦T cells is the paucity of the identified ligands for these cells. To date, only a few human TCR␥␦-recognized protein antigens have been reported. To better understand ␥␦T cell function, we have developed a new technical strategy to identify human TCR␥␦-recognized epitopes/proteins by combining immunological and biochemical methods using the CDR3 peptide of T cell receptor ␦2 chain (CDR3␦)-mediated epitope/protein-binding assays (CEPAs). This CDR3␦ peptide-based immunobiochemical strategy was based on a hypothesis that the primary sequence of CDR3, especially CDR3␦, because of its similarity to CDR3␤ of TCR␤ and V H CDR3 of BCR in gene composition, could serve as the key determinant of specificity in antigen binding by the TCR␥␦. We have confirmed this hypothesis using in vitro binding assays. Synthesized TCR V␦2 CDR3 peptides derived from tumor-infiltrating lymphocytes (TILs) in ovarian epithelial carcinoma (OEC) could specifically bind tumor cell lines and tissues, suggesting the determinant role of CDR3␦ in antigen binding. Moreover, CDR3␦ peptide-mediated binding specificity was blocked by preincubation with the same peptide, which decreased the cytotoxicity of ␥␦T cells to OEC cells in vitro, further indicating such binding is specific (21). Based on our previous findings, we then used synthesized CDR3␦ peptides as specific probes to identify putative TCR␥␦-recognized antigenic epitopes in a peptide library and TCR␥␦-recognized proteins in an affinity chromatography system. As shown in supplemental Fig. 1, CDR3␦ peptide-based immunobiochemical strategy is technically composed of four steps as follows: selecting probes for epitopes/proteins, screening epitopes/proteins, identifying epitopes/proteins sequences, and validating epitopes/proteins functions. Based on the specific binding of CDR3␦ peptide to its target, we developed a set of CEPAs, which contains eight assays running through all the four steps of our strategy, including the following: 1) CDR3␦ peptide-mediated surface plasmon resonance (SPR) to determine interaction between CDR3␦ peptides and target proteins; 2) CDR3␦ peptide-mediated immunofluorescence assays to detect specific binding of CDR3␦ peptides to target cells; 3) CDR3␦ peptide-mediated enzyme immunoassays to detect specific binding of CDR3␦ peptides to target proteins and tissues; 4) competitive test to confirm CDR3␦ peptide sharing antigenic binding specificity with TCR␥␦; 5) panning putative epitope peptides in a 12-mer random peptide phage-displayed library with immobilized CDR3␦ peptides; 6) screening antigenic proteins in affinity chromatography with CDR3␦ peptide as affinity molecule; 7) phage-ELISA to select single CDR3␦ peptide-binding phage clone; and 8) CDR3␦ peptide-mediated Western blotting to detect CDR3␦ peptide-binding proteins.
With this strategy, we have identified seven tumor-related epitopes and two HBV infection-related epitopes, as well as two self-proteins, including HSP60 and human mutS homolog 2 (hMSH2). HSP60 has been identified as a ligand for ␥␦T cells (18,19), whereas hMSH2 is a DNA mismatch repair protein in the cell nucleolus (22). We show that hMSH2 is expressed on tumor cell membrane and recognized by V␦2 ␥␦T cells, suggesting that hMSH2 may be a novel ligand for TCR␥␦. Taken together, our findings provide a novel strategy to identify epitopes/proteins for TCR␥␦.

EXPERIMENTAL PROCEDURES
Peptide Synthesis-The following four V␦2 CDR3 peptides were synthesized, including three (OT1, OT2, and OT3) corresponding to the sequences derived from ␥␦TIL in OEC and one CDR3 peptide HP1, whose sequence was derived from a peripheral ␥␦T cell clone from an HBV-infected patient: OT1, CACDSHGPSRLMMEGGLLGTDKLIFGKG; OT2, CARKDLPINNWGIPRIDKLIFGKG; OT3, CDFPSHTFHS-TGGHTTDKLIFGKG; and HP1, CACDHLPLGDTRVHDK-LIFGKG. Meanwhile, a structurally very similar peptide HP1m was synthesized with mutations of the first four amino acids and was employed as the control for HP1: HP1m, TQEHHLPL-GDTRVHDKLIFGKG. The conformational preference of the CDR3␦ peptides OT3 and HP1 was analyzed by molecular dynamics. The results suggested that both of them displayed loop-like conformations in water (supplemental Fig. 2 shows the C-␣ root mean square deviation of the CDR3␦ peptides). The modeled structure of OT3 and HP1 is shown in Fig. 1A and Fig. 2A, respectively. Seven tumor-related epitope peptides, named as EP1 to EP7, and two HBV infection-related epitopes, EP8 and EP9, were also synthesized ( Table 1). All peptides were biotinylated by adding a biotin tag at the N terminus of each peptide sequence for detection in the following assays. Peptides were synthesized by the Academy of Military Medical Sciences, China. The purity of each peptide was 85% in high performance liquid chromatography analysis.
Cell Culture and ␥␦T Cell Preparation-A human OEC cell line SKOV3 was cultured in McCoy's 5A medium (HyClone) supplemented with 10% fetal calf serum (FCS), a gift from Dr. Keng Shen (Department of Gynecology, Peking Union Medical College Hospital, Beijing, China). Another human OEC cell line HO8910 was established in our laboratory and cultured in RPMI 1640 medium (Invitrogen) with 10% FCS. A human uterine cervix cancer cell line (HeLa) and a human T lymphoma cell line (J.RT3-T3.5) deficient in both TCR ␣ and ␤ chains were obtained from the American Type Culture Collection (ATCC) and maintained in 10% FCS/Dulbecco's modified Eagle's medium (Invitrogen) and 10% FCS/RPMI 1640 medium (Invitrogen), respectively. Human liver cancer cell line HepG2 and an HBV whole genome-transfected human liver cancer cell line (2.2.15 cell) were obtained from Cell Bank at the Institute of Basic Medical Sciences, Chinese Academy of Medical Sciences, China, and cultured with 10% FCS/minimum Eagle's medium/ NEAA medium (Invitrogen) and 10% FCS/minimum Eagle's medium (Invitrogen), respectively. Fresh PBMC separated from peripheral blood of healthy donors by density gradient centrifugation on Ficoll-Hypaque (GE Healthcare) were grown in RPMI 1640 medium (Invitrogen) with 10% FCS/IL-2 (200 units/ml) in 24-well culture plates with immobilized anti-pan-TCR␥␦-monoclonal antibody (Immunotech). After 2 weeks of culture, the purity of ␥␦T cells was Ͼ90% as assessed by flow cytometry analysis and ready for further use.
SPR Assay-SPR experiment was performed using an IAsys biosensor (Affinity Sensors). The surface of CMD chip was activated by 200 mM N-ethyl-NЈ-dimethylaminopropylcarbodiimide and 50 mM N-hydroxysuccinimide. 10 ng of HSP60 was coupled to the activated CMD chip. After immobilization, the surfaces were blocked with 1 mM ethanolamine (pH 8.5) and then washed with 1 mM formic acid to remove noncoupled proteins. OT3 peptide from low to high concentrations (0.25 to 2 mM) consecutively and HP1/HP1m peptides (0.5 mM) were employed for the binding assay at room temperature. BIAevaluation version 3.1 (Biacore AB) was used for data analysis.
FCM and Confocal Microscopy Assays-Immunofluorescence assays were applied to assess CDR3␦ peptide/OT3 graft immunoglobulin (OT3-Ig), whose V H CDR3 region of a human autoantibody (IgG) recognizing thyroglobulin was replaced with OT3 peptide sequence (21), epitope peptide-mediated binding activity to target cells (tumor cells/␥␦T cells), and the expression of hMSH2 on tumor cells. For FCM assay, cells were incubated with biotinylated CDR3␦/epitope peptides, OT3-Ig (mouse IgG), and rabbit anti-hMSH2 polyclonal antibody (Santa Cruz Biotechnology). FITC-conjugated streptavidin (Pierce) or FITC-conjugated goat anti-mouse/rabbit IgG antibody (Pierce) was then added and incubated for 30 min at 4°C. Wide type immunoglobulin (WT-Ig) and rabbit IgG (Zhongshan, China) were used as controls for OT3-Ig and rabbit anti-hMSH2 polyclonal antibody. The cells were analyzed on a FACSort flow cytometer (BD Biosciences). For confocal microscopy, cells were fixed on slides by 2% cold paraformaldehyde and in turn incubated with biotinylated synthesized HP1 peptide and with FITC-conjugated streptavidin. Cells incubated with biotinylated synthesized HP1m peptide instead of the HP1 peptide were used as controls. Slides were examined with a confocal laser microscope (LSM 510; Carl Zeiss).
Immunohistochemistry and ELISA-Immunohistochemistry assay was used to evaluate CDR3␦ peptide-mediated binding activity to target tissues. Briefly, formalin-fixed paraffin-embedded sections of HBV-infected liver tissues or normal tissues were deparaffinized and then boiled by microwave for antigen retrieval. After quenching with hydrogen peroxide, the sections were blocked with 5% goat serum. Biotinylated synthesized HP1 peptide (5 g) was then added to the slides. The sections were subsequently incubated with horseradish peroxidase (HRP)-conjugated streptavidin (Zhongshan, China). Binding was visualized using diaminobenzidine (Sigma) as the substrate and observed under a light microscope. In CDR3␦-ELISA, 96-well plates were coated with proteins extracted from SKOV3 cells (10 g/well) in 0.1 M NaHCO 3 (pH 9.6), or formalin-fixed SKOV3 cells (4 ϫ 10 5 /well). After blocking with 3% bovine serum albumin solution, the plates were incubated with biotinylated CDR3␦ peptides or OT3-Ig/WT-Ig. After reaction with HRP-conjugated streptavidin (Pierce) or HRP-conjugated goat anti-mouse IgG antibody (Pierce) and substrate (Sigma), the plates were then read on a microplate reader (Labsystem) at 450 nm.
MTT Colorimetric Test and CDR3␦ Peptide Competitive Test-This test was used for evaluation of ␥␦T cell cytotoxicity to tumor lines. Briefly, SKOV3 cells as target cells were seeded onto 96-well plates (1 ϫ 10 4 /well). For competitive testing, target cells were preincubated with CDR3␦ peptides (0.1 mg/ml) or anti-hMSH2 polyclonal antibody (0.2 mg/ml) at 37°C for 2 h, and equivalent rabbit IgGs (Zhongshan, China) were used as controls for anti-hMSH2 polyclonal antibody. Then ␥␦T cells as effector cells were incubated with target cells for another 8 h (the ␥␦T cells were preincubated with epitope peptides for 2 h in stimulating assay). MTT solution (5 mg/ml) was added to wells (15 l/well) and incubated at 37°C for 4 additional hours. The reaction was stopped by the addition of 100 l of lysis buffer (10% SDS) to dissolve the tetrazolium crystals. The plate was examined in a Multiskan Microplate Reader (Thermo Labsystems, Finland) and the percentage of specific lysis was calculated by using Equation 1, T A ϭ absorbance of target cells for control; E A ϭ absorbance of effector cells for control, and (E ϩ T) A ϭ absorbance of effector cells cultured with target cells. The absorbance was determined at 570 and 630 nm.
Panning of CDR3␦ Peptide-binding Phage Clones in a 12-mer Random Peptide Phage-displayed Library-A 12-mer random peptide phage-displayed library (New England Biolabs) was screened by CDR3␦ peptides as follows: a 96-well plate was coated with 100 l/well coating buffer (containing 10 g CDR3␦ peptide) for 2 h at 37°C and then blocked with PBS containing 3% bovine serum albumin overnight at 4°C. The primary library solution was added to the wells (10 l per well containing 10 11 colony-forming units) and shaken gently at room temperature for 1 h. After thoroughly washing with PBS, 0.1% Tween 20, the CDR3␦ peptide-binding phages were eluted by acidic buffer (0.2 M glycine-HCl (pH 2.2)) and neutralized by 1 M Tris-HCl (pH 9.1) immediately. In some cases, a high concentration of CDR3␦ peptide (200 g/ml) was used for compet-itive elution as well. The concentration of Tween 20 in washing buffer was increased to 0.3 and 0.5% in the following two rounds of biopanning and was kept at 0.5% thereafter.
Affinity Chromatography-Affinity chromatography was used to enrich CDR3␦ peptide-binding proteins with the following procedures. CDR3␦ peptide was coupled to a 1-ml HiTrap N-hydroxysuccinimide-activated HP column (Amersham Biosciences) (5 mg) through its C-terminal hydroxyl group in coupling buffer (0.2 M NaHCO 3 , 0.5 M NaCl (pH 8.3)). The excess active groups uncoupled to CDR3␦ peptide were deactivated with solution (0.5 M ethanolamine, 0.5 M NaCl (pH 8.3)). CDR3␦ peptide-coupled column was then thoroughly equilibrated by binding buffer (20 mM PBS (pH 7.0)). After the proteins were applied to the column, the absorbance value of the outflow at 280 nm was detected, and elution buffer (1% formic acid) was changed until the A 280 returned to the base line. The fractions containing eluted proteins were collected and neutralized quickly with 1 M Tris-HCl (pH 9.1). Throughout the process, the flow rate was kept under 0.2 ml per min.
Identification for CDR3␦ Peptide-binding Phage Clones by Phage-ELISA, PCR, and Sequencing-After five rounds of panning, individual plaque was picked randomly, and positive phage clones were selected by phage-ELISA using HRP-anti-M13 phage antibody (Amersham Biosciences). Briefly, wells of ELISA plate were coated with CDR3␦ peptides and blocked as in the panning procedures. Single colony phage was added into CDR3␦ peptide-coated wells and incubated at 37°C for 2 h. After fully washing, HRP-anti-M13phage antibody was added and incubated at 37°C for 1 h. The positive phage clones in ELISA were then amplified, and epitopes containing parts were obtained by PCR and sequenced with the ABI automatic sequencer 377 (PerkinElmer Life Sciences) using the Ϫ96 sequencing primer, 5Ј-CC CTC ATA GTT AGC GTA ACG-3Ј (supplied in the kit).
Identification of CDR3␦ Peptide-binding Proteins by SDS-PAGE and CDR3␦ Peptide-mediated Western Blotting-The enriched proteins through affinity chromatography were separated by SDS-PAGE. After proteins were electrically transferred to a nitrocellulose membrane, the blot membrane was placed in blocking buffer. The biotinylated CDR3␦ peptide (10 g/ml) was used as the primary antibody, and HRP-streptavidin was employed as the secondary antibody for usual Western blotting. Chemiluminescent HRP substrate (Pierce) was added, and the blot was exposed to a x-ray film for an appropriate duration.
Data Base Search and Protein Identification-The in-gel trypsin digestion and MS analysis of a certain SDS-PAGE protein band was carried out at Beijing Proteome Research Center, China. LC ESI-MS/MS analyses were performed using an LTQ system (Thermo Finnigan). Tryptic peptides mixtures were first loaded onto a capillary column (15 cm length, 75 m inner diameters). The elution from reverse phase liquid chromatography column was directed on line to be detected by MS. All the MS/MS spectra were searched using SEQUEST algorithmbased Bioworks version 3.2 (Thermo Finnigan, San Jose, CA) against the International Protein Index human data base 3.19, which had 60,397 entries (23). Proteins with at least two unique "identity" score peptides were considered as being unambigu-ously identified, and the single peptide-matched proteins were renounced.
Epitope Peptide-immobilized Expanding Assay and [ 3 H]TdR Incorporation Assay-For expanding assay, PBMC were cultured in 24-well plates with immobilized epitope peptides (10 g/well), using blank wells and anti-pan-TCR␥␦-monoclonal antibody (1 g/well)-coated wells as negative and positive controls, respectively. Cells were then cultured in RPMI 1640 medium (Invitrogen) with 10% FCS and IL-2 (200 units/ml). The percentage of ␥␦T cells was measured 2 weeks later by FCM. In [ 3 H]TdR incorporation assay, ␥␦T cells (purity Ͼ90%) were first cultured without IL-2 for 48 h. The resting ␥␦T cells were then seeded onto 96-well plates with 1 ϫ 10 5 cells per well. After 60 h of stimulation with epitope peptides or IL-2 plus epitope peptides, [ 3 H]TdR was added into wells, and the cells were cultured for an additional 8 h. The proliferation of ␥␦T cells was measured by counts/min value.

RESULTS
Specific Binding Activity of Synthesized CDR3␦ Peptides to OEC Cell Line SKOV3-To work as probes, synthesized CDR3␦ peptides must bind specifically to tumor cells. Although we previously reported the specific binding activities (21), several binding experiments were repeated to confirm the binding. Fig.  1B showed that the peptide OT3 bound to tumor cell lines but not to normal PBMC. OT3-mediated ELISA was used for evaluation of the interactions between OT3 and tumor cells. Both SKOV3 cells (Fig. 1C) and SKOV3 total protein extracts (Fig.  1D) bound OT3 peptide in a dose-dependent manner. OT3 competed with TCR␥␦ on ␥␦T cells expanded from PBMC (Fig.  1E), resulting in a reduced cytotoxicity of human ␥␦T cells to SKOV3 cells in vitro (Fig. 1F). These data suggest that OT3 peptide and TCR␥␦ bind the same ligands on SKOV3 cells. Similar results have been obtained from peptides OT1-and OT2-mediated binding and competitive assays (data not shown). The interaction between the OT3 graft immunoglobulin (OT3-Ig) and tumor cell lines/protein extracts were analyzed by FCM and ELISA, respectively. Similar to OT3 peptide, the OT3-Ig bound to SKOV3 and HeLa, whereas the WT-Ig did not bind to any cell lines (Fig. 1G). OT3-Ig bound to SKOV3 total protein extracts in ELISA, and the WT-Ig did not (Fig. 1H), suggesting specific binding.
CDR3␦ Peptide HP1 Specifically Binds to HBV-infected Tissues/Cells-To further investigate CDR3␦ peptide binding specificities, we analyzed the binding of HP1 peptide to HBV-infected liver tissues and cell line 2.2.15 containing the intact dimers of the HBV genome, respectively. HP1 peptide specifically bound 2.2.15 cells but not normal cells (PBMC) and other types of tumor cell lines, such as SKOV3 and HepG2 (Fig. 2B). In immunohistochemistry assay, HP1 peptide specifically bound hepatocarcinoma tissue (Fig. 2C). Strong fluorescent staining was detected on HBV-infected 2.2.15 cells after incubating with HP1 peptide in confocal microscopy assay (Fig. 2D). Consistent with the above results, ELISA shows the specific interaction of HP1 peptide and HBV-infected 2.2.15 total protein extract (Fig. 2E). Taken together, our data suggest that the binding of synthesized HP1 peptides to targets, either tissue, cell, or protein extract, is specific.
Biopanning for Epitopes from Phage Library Using CDR3␦ Peptides-To identify epitopes recognized by CDR3␦ peptides, a CDR3␦ peptide-mediated biopanning of a 12-mer random peptide phage-displayed library was performed. The positive phage population was enriched in five rounds of panning, with output to input ratio about 1 ϫ 10 Ϫ3 (supplemental Table 1). Through the selection of phage-ELISA, 120 phage clones binding specifically to immobilized-CDR3␦ peptides were sequenced. As shown in supplemental Table 2, the sequences of displaying peptide on positive phage clones revealed that these clones share several motifs. We chose seven OEC tumor-related sequences and two HBV infection-related sequences as candidate epitopes from the sequenced phages based on their high frequencies among all 120 sequences. These peptide epitopes were chemically synthesized (named EP1-9) for functional assays (Table 1).
BLAST-BLAST search was performed to identify human proteins containing motifs in the sequences of the CDR3␦ peptide-bound phage displaying peptides. The BLAST results of the nine sequences that we chose as candi-FIGURE 1. Synthesized CDR3␦ peptide OT3 specifically bound tumor cell line SKOV3. A, representative structure of OT3 peptide in molecular dynamics simulation. C-␣ trace was shown with residues colored according to their types (blue for acidic amino acids, red for basic, yellow for polar, and gray for nonpolar residues). B, flow cytometry analysis of surface staining of tumor cell lines by OT3 peptide. Normal PBMC was used as control. Cells were stained with biotinylated (bio)-OT3 peptide and FITC-conjugated streptavidin. C and D, ELISA of OT3 peptide binding to SKOV3 cell and its protein extracts. SKOV3 cells (4 ϫ 10 5 /well) or protein extracts (10 g/well) have been immobilized on ELISA plate. The bio-OT3 peptide was added with various concentrations, and HRP-conjugated streptavidin was added subsequently. Each assay was run in duplicate. *, p Ͻ 0.05, A value of SKOV3 cell-coated wells detected with bio-OT3 (1 g/well), when compared with that of mock wells; **, p Ͻ 0.01, A value of SKOV3 cell coated wells detected with bio-OT3 peptide (10 g/well), when compared with that of mock wells; A value of SKOV3 protein extracts coated wells detected with bio-OT3 peptide (1 or 10 g/well) compared with that of mock wells. E and F, OT3 peptide competitive binds SKOV3 in vitro. MTT assay evaluating cytotoxicity inhibition of ␥␦T cells against SKOV3 after OT3 peptide preincubation. PBMC-derived ␥␦T cells with the purity more than 90% in FCM analysis were used as effector cells in MTT assay (E). OT3 peptide (10 g/well) was incubated with SKOV3 cells for 2 h before adding ␥␦T cells. Results are representative of three independent experiments (F). The A value of OT3 peptide added wells was compared with that of wells without OT3 (*, p Ͻ 0.05, E:T ratio ϭ 10:1; **, p Ͻ 0.01, E:T ratio ϭ 20:1, and E:T ϭ 40:1). E:T stands for the ratio of effector cells to target cells. G, flow cytometry analysis of surface staining by OT3-graft Ig with tumor cells SKOV3 and HeLa (blue). WT-Ig was used as control (red). Results are representative of three independent experiments. H, ELISA of OT3-graft Ig to SKOV3 total protein extract. The proteins extracts have been coated on an ELISA plate (1 g protein/well). The OT3-Ig and WT-Ig were added, respectively, to the plate in the same concentration (5 g/ml). Each assay was run in duplicate. date epitopes are listed in supplemental Table 3. No exactly matched protein sequences were returned from BLAST analysis, which was based on primary sequence searching, and the best matching manner had a maximum of 7 uninterrupted amino acids matching all 12 amino acids. In addition, most proteins in BLAST results were from prokaryotes or insects and hypothetical proteins, consistent with a role of TCR␥␦ in innate immune recognition. The BLAST results of other CDR3␦ peptide-bound motifs were similar (data not shown).
Epitope Peptides Bind and Activate ␥␦T Cells in Vitro-To test whether the synthesized putative epitope peptides specifically bind TCR␥␦, we performed a binding assay using FCM. As shown in Fig.  3A, 65% of ␥␦T cells showed a binding activity to biotinylated peptide EP7, whereas J.RT3-T3.5 cells, a T cell line deficient in TCR chains, had much lower binding percentage. All immobilized peptides EP1-6 were able to induce proliferation of ␥␦T cells in PBMC to an average 16% in 2 weeks, markedly higher than negative control (2.5%). The percentages of expanded ␥␦T cells in PBMC ranged from 7.8% (EP1 and EP5) to 32.3% (EP3) (Fig. 3B). We determined the sequences of CDR3 regions in both ␥9 and ␦2 chains of EP-expanded ␥␦T cells and found that EP-expanded TCR␥␦ cells were polyclonal, but with some common motifs in the CDR3␦ region (supplemental Table 5). The proliferation of ␥␦T cells induced by soluble peptides EP3-5 further confirmed such stimulating activities of epitope peptides. Significant proliferation of ␥␦T cells to peptides EP3/4 (Fig. 3C) and EP5 (Fig. 3D) was observed in a dose-dependent manner. Peptide EP6 induced the cytotoxicity of ␥␦T cells to SKOV3 in a dose-dependent manner (Fig. 3E). Taken together, these data demonstrate that these epitope peptides not only bind to but also activate ␥␦T cells. Similarly, HBV infection-related epitope peptides EP8 and EP9 had similar

. HBV infection-related CDR3␦ peptide HP1 specifically bound HBV-infected tissues/cells/cell extracts, and HBV infection-related epitope peptides stimulated ␥␦T cells in vitro.
A, representative structure of HP1 peptide in the molecular dynamics simulation. C-␣ trace was shown with residues colored according to their type (the same as in Fig. 1A). B, flow cytometry analysis of surface staining by HP1 peptide to PBMC, tumor cell lines (SKOV3 and HepG2), and HBV-infected cells 2.2.15 (red). For cells with HP1m peptide, a structurally very similar peptide to HP1 with mutations of the first four amino acids, staining served as negative controls (green). Blue line shows the staining with PBS. C, immunohistochemistry staining of hepatocarcinoma tissues. The same tissues were stained with biotinylated HP1 peptide, HP1m peptide, and PBS as shown from left to right. Staining was visualized using diaminobenzidine as the substrate (brown, ϫ200). D, confocal microscopy analysis of the staining of HP1 peptide on HBV-infected cells 2.2.15. Binding of biotinylated HP1 peptide was analyzed by confocal microscopy after staining with FITC-conjugated streptavidin. As a control, control peptide HP1m and PBS instead of HP1 peptide was used. E, HP1 peptide specifically bound 2.2.15 protein extracts in ELISA. **, p Ͻ 0.01, A value of 2.2.15 cell protein extract-coated wells detected with HP1 peptide, when compared with all controls, including blank, HepG2 protein extracts, and HP1m peptide. F, immobilized EP8 and EP9 could induce expansion of ␥␦T cells from PBMC. We used panel 1, RPMI 1640 medium mock as negative control; panel 2, immobilized 1 g of anti-pan-TCR␥␦ monoclonal antibody as positive control; panels 3 and 4, immobilized 10 g of epitope peptides EP8, EP9 respectively. G, EP8 and EP9 could induce proliferation of resting ␥␦T cells at 10 g/ml in [ 3 H]TdR incorporation assay. effects on ␥␦T cells in inducing expansion of ␥␦T cells in PBMC (Fig. 2, E and F).
Identification of OT3-specific Tumor Proteins-We screened the total cell lysate of SKOV3 tumor cells directly using OT3mediated affinity chromatography. The 1-ml HiTrap N-hydroxysuccinimide-activated HP column coupled with 4.5 mg of OT3 with a coupling ratio of 90% (data not shown). After washing with elution buffer, the elution peak was collected (Fig. 4A). The protein fractions were pooled and concentrated. SDS-PAGE showed that only two protein bands with molecular mass of 45 and 60 kDa, respectively, can be obviously observed (Fig.  4B). Western blotting analysis indicates that only the 60-kDa protein was OT3-specific, and its amount was enriched when compared with SKOV3 total proteins (Fig. 4C). The specific band in the SDS-polyacrylamide gel was analyzed by LC ESI-MS/MS. A total of 15 proteins with at least two unique peptides was identified for the 60-kDa protein under the appropriate cutoff parameters for SEQUEST search, among which pyruvate kinase 3 isoform 1, DNA mismatch repair protein (hMSH2), and HSP60 demonstrated highly matching unique peptides count and higher cover percentage (supplemental Table 4).
Validation of HSP60 and hMSH2-Stressful conditions upregulate the expression of HSP60, which has been considered as a target antigen involved in the activation of ␥␦T cells (18,19). We examined the binding of CDR3␦ peptide OT3 to HSP60 using SPR. OT3 peptide indeed bound HSP60 in a dose-dependent manner (supplemental Fig. 3). We chose hMSH2 as a potential new ligand for TCR␥␦ for further characterization. In FCM, 21% of HepG2 and 43% of SKOV3 tumor cell lines were positive for surface staining with antibody against hMSH2, whereas less than 7% of normal ␥␦T cells expressed hMSH2 (Fig. 4D). In cytotoxicity blocking assay, cytotoxicity of V␦2 ␥␦T cells to SKOV3 cells was blocked by anti-hMSH2 antibody, whereas cytolytic activity of V␦1 ␥␦T cells could not be inhibited (Fig. 4, E and F), suggesting that hMSH2 may be a ligand for V␦2 ␥␦T cells. In ongoing investigations, we are cloning and expressing recombinant hMSH2 protein for further functional tests. Preliminary data revealed that hMSH2 cloned from SKOV3 cDNA was mutated (data not shown), suggesting that the mutant hMSH2 might be displayed on tumor cell surface to be recognized by immune cells.

DISCUSSION
In this study, we demonstrate a novel strategy to identify epitopes/proteins recognized by ␥␦T cells by combining immunological and biochemical methods. With this strategy, we identified nine ␥␦T cell-reactive epitopes and two antigenic proteins, HSP60 and hMSH2. Another putative antigenic protein (pyruvate kinase 3 isoform 1) showed possible sequence characteristics as a ligand for TCR␥␦ according to MS analysis, although the validation proceedings continue. Therefore, our method provides a novel and specific high throughput system to identify epitopes/proteins recognized by ␥␦T cells.
This strategy includes four independent but closely related technical steps, including selecting probe for epitopes/proteins, screening epitopes/proteins, identifying epitope/protein sequences, and validating epitope/protein functions. The binding specificity of synthetic CDR3␦ peptides to target molecules forms a principal base for a set of CEPAs.
In selecting probes for ␥␦T cell-recognized epitopes/proteins, we used four assays to verify the specificity of these peptides, including CDR3␦ peptide-mediated SPR, enzyme immunoassay, immunofluorescence assays, and peptide competition tests. Biotin-streptavidin affinity system makes an amplification signal, which includes enzyme immunoassays as well as immunofluorescence assays. These results were consistent with the nonbiotinylated CDR3␦ peptide in the SPR assay, indicating the CDR3␦ peptides were specific and could be used as selecting probes. Moreover, data from the CDR3␦ peptide competitive test further confirmed the specificity of CDR3␦ peptide for their targets. Although CDR3␦ peptide showed good binding activities to targets, there were some shortcomings. A linear peptide with less than 30 amino acid residues would have some unspecific binding. To circumvent this, we constructed and expressed a CDR3␦-graft antibody, whose CDR3 sequences in heavy chains were replaced by CDR3␦. The CDR3␦-graft antibody showed the similar binding specificity with CDR3␦ peptides. In summary, the tested CDR3␦ peptides showed specific binding activities to target proteins, cells, and tissues and can serve as probes for screening epitopes/proteins. CDR3␦ peptides were employed in two different ways: panning a 12-mer random peptide phage-displayed library and There are two biopanning manners including nonspecific (NS) and specific (S). NS is nonspecific elution by acidic buffer, and S stands for specific elution by high dose of corresponding probe. b Frequency of predominant clone expressing same peptide in all tested phage clones was listed in supplemental Table 2. For example, 9/51 means this peptide appears 9 times in all 51 sequenced phage clones. c EP2 sequence (PLLPMHPMKVSH) was found in one clone of all 35 tested phage clones with varied sequences of peptides, 10 of which shared conserved motif LPMHPM.
binding by affinity chromatography. One issue is whether we can obtain possible epitope peptides in the 12-peptide library. It is reported that T22, an identified protein recognized by murine ␥␦T cells, formed a TCR␥␦-recognized epitope 15-18 amino acid residues in length (24). Therefore, the 12-mer random peptide phage-displayed library might cover most parts of the TCR␥␦-recognized epitopes, suggesting that the strategy is feasible. We used four CDR3␦ peptide probes in this study. As a result, we found that CDR3␦ peptide had sufficient affinity to its bound proteins in the CDR3␦ peptide-mediated affinity chromatography, which was confirmed by the enriched collection of eluted proteins from SKOV3 cell total proteins with an elution peak. Therefore, our data suggest that CDR3␦ peptidemediated affinity chromatography is useful for isolation of ligand for TCR␥␦. Because the principle of this technique is based on interaction between the ligand-binding side (such as CDR3) and the receptor-binding side (such as epitope), the peptides from the ligand-binding side in other immune receptors, such as NKG2D, TLR, and BCR, might also be tested for their ligand proteins in a similar manner.
We have characterized nine peptides as putative epitopes, among which seven were tumor-related and two were HBV infection-related. BLAST search revealed that most matched proteins were conserved proteins of prokaryotes. We did not find any related human proteins in BLAST matching these putative epitope peptides, suggesting the interaction between TCR␥␦ and ligands, epitopes may have three-dimensional conformation rather than linear structures. Moreover, we used LC-ESI MS/MS to identify the sequences of putative antigenic proteins after CDR3␦ peptide-mediated Western blotting. We found that 3 proteins among 54 are considered to be possible ligands for TCR␥␦, including pyruvate kinase 3 isoform 1, hMSH2, and HSP60. An important feature of our strategy was that the proteomic technique was used in combination with the immunological/biochemical method. On the one hand, the specificity of the immunological/ biochemical method provides reliable support to the proteomic analysis, resulting in a high probability of successfully identifying TCR␥␦ recognized proteins. On the other hand, proteomic technique brings an obvious advantage to this strategy, i.e. rapid and high throughput.
In validating epitopes/proteins functions, we tested the binding of putative epitopes to ␥␦T cells through a series of experiments. These CDR3␦-bound epitope peptides were able to bind ␥␦T cells and could trigger their proliferation and cytotoxicity to target cells in vitro. FIGURE 3. Tumor-related epitope peptides bound and stimulated ␥␦T cells in vitro. A, flow cytometry analysis of ␥␦T cells surface staining by biotinylated epitope peptide EP7. The same staining of J.RT3-T3.5 cells was performed as control group. B, expansion of ␥␦T cells from PBMC by immobilized epitope peptides. Panel 1, RPMI 1640 medium was added as negative control; panel 2, 1 g of anti-pan-TCR␥␦ monoclonal antibody was immobilized as positive control; panels 3-8, 10 g of epitope peptides EP1-6 were immobilized, respectively. The figure shows phycoerythrin (PE)-anti-TCR␣␤ and FITC-anti-TCR␥␦ double staining after culturing for 2 weeks. C and D, epitope peptides EP3, EP4, and EP5 triggered the proliferation of resting ␥␦T cells in [ 3 H]TdR incorporation assay. The counts/min value showed the activation status. E, MTT assay to evaluate the change of cytolytic activity of the ␥␦T cells against SKOV3 on the binding of EP6. ␥␦T cells expanded from PBMC as mentioned above were used as effector cells, and SKOV3 were target cells. Different doses of EP6 were incubated with ␥␦T cells for 2 h before adding to target cells. Results are representative of three independent experiments. could be blocked by anti-hMSH2 antibody. It suggests that ectopically expressed hMSH2 in malignant situation might become a novel ligand for V␦2 ␥␦T cells. The ectopic expression of hMSH2 in transformed cells may alert ␥␦T cells of the transformed cells. In summary, we successfully obtained functional epitopes/proteins for TCR␥␦ through a novel strategy, based on the core technique CEPA, demonstrating that it is useful in identification of ␥␦T cell-recognized epitopes/proteins.