Int. J. Cancer: 76, 304?311 (1998) r 1998 Wiley-Liss, Inc. Publication of the International Union Against Cancer Publication de l?Union Internationale Contre le Cancer TUMOR-SPECIFIC PEPTIDES IN CUTANEOUS T-CELL LYMPHOMA: ASSOCIATION WITH CLASS I MAJOR HISTOCOMPATIBILITY COMPLEX AND POSSIBLE DERIVATION FROM THE CLONOTYPIC T-CELL RECEPTOR Carole L. BERGER1*, B. Jack LONGLEY2, Suguru IMAEDA1, Inger CHRISTENSEN1, Peter HEALD1 and Richard L. EDELSON1 of Dermatology, Yale University School of Medicine, New Haven, CT, USA 2Department of Dermatology, Section of Dermatopathology, Columbia University, College of Physicians and Surgeons, New York, NY, USA 1Department We wished to identify and characterize tumor-associated class I peptides which could potentially serve as immunogens for an immunoprotective CD8 response in cutaneous T-cell lymphoma (CTCL). Candidate idiotypic peptides were identified from the third complementarity determining region (CDR3) of the clonotypic T-cell receptor (TCR) expressed on malignant T cells and native class I peptides were identified from CTCL cells. Idiotypic peptides were designed by sequencing of patients? CDR3 and identifying 9 amino acid peptides that could be accommodated in the peptide-binding motif of the class I alleles. Three candidate idiotypic peptides were synthesized and tested by measuring release of tumor necrosis factor-a (TNF-a) from autologous CD8 cells. Native peptides were acid-eluted from class I molecules on CTCL lymphocytes, fractionated, tested in the TNF-a assay and sequenced. Two unique idiotypic peptides were specifically recognized by autologous CD8 cells from CTCL patients. In addition, a native peptide eluted from class I molecules of CTCL tumor cells was identified, in the protein data base, as a novel molecule with partial sequence homology to the conserved portion of the patient?s TCR. This homology was used to construct an extended native peptide sequence that was immunogenic for CD8 cells from both CTCL patients. Our results demonstrate that peptides derived from the TCR can be used as tumor-specific immunogens that are recognized by CD8 cells. Moreover, novel class I peptides isolated from the tumor cell also serve as immunogens. These peptides might form the basis of an anti-tumor vaccine for immunotherapy of CTCL. Int. J. Cancer 76:304?311, 1998. r 1998 Wiley-Liss, Inc. The goal of tumor immunotherapy is selective recognition and destruction of malignant cells. For this to occur, distinctive antigenic peptides must be expressed on the tumor cell surface in association with the major histocompatibility complex (MHC) class I or II molecules for presentation to CD4 and CD8 T lymphocytes. Malignant cells of cutaneous T-cell lymphoma (CTCL), a clonal proliferation of CD41 (Edelson et al., 1979; Kung et al., 1981), memory T cells (Picker et al., 1990) with an initial affinity for epidermis, display tumor-specific antigens, which can be recognized by autologous CD8 T cells (Berger et al., 1996). Suggestions that these tumor antigens can be targeted in vivo derives from 2 sets of data. First, initial malignant infiltration of skin is often associated with a pronounced CD8 host response (Bagot et al., 1992), which diminishes as the disease disseminates, and the intensity of this CD8 response correlates with a good prognosis (Wood et al., 1991). Second, remissions in leukemic CTCL patients treated with photopheresis (Edelson et al., 1987), thought to immunize patients against malignant cells (Edelson et al., 1994), appear to require an intact CD8 T-cell compartment (Heald et al., 1992). Tumor-specific antigens may be derived from viral gene products, point mutations of normal proteins and normal differentiation antigens preferentially displayed on the malignant cells. A potential source of immunogenic unique peptides in T-cell tumors is the clone-specific T-cell receptor (TCR). Anti-tumor immune responses directed toward the idiotypic components of the CTCL TCR would solely impact the malignant T-cell clone. Therefore, we have investigated the possibility that the class I MHC-associated clone-specific peptides identified on CTCL cells may originate, at least in part, from degradation products of unique components of the TCR protein. We report here our findings that peptide components of the idiotypic region of the CTCL TCR beta variable chain (Vb) are selectively recognized by autologous CD8 T cells and that a novel native peptide directly extracted from the CTCL tumor cells is also a target for autologous CD8 T cells. Together, these findings indicate that a single clone of CTCL cells is characterized by multiple class I-associated antigenic peptides, permitting an additive anti-tumor attack. These observations suggest that it may be possible to construct peptide vaccines, based on the sequence of the TCR, for tumor immunotherapy of CTCL. MATERIAL AND METHODS Patient population Leukapheresis blood was obtained from CTCL patients undergoing therapeutic photopheresis treatment in accordance with the Yale University Human Investigational Review Board guidelines. The 2 patients studied had a clonal expansion of malignant T cells that expressed a Vb8a TCR identified by flow cytometry using a family-specific monoclonal antibody (MAb) (Table I; T Cell Sciences, Cambridge, MA). The class I histocompatibility locus alleles of the patients? peripheral blood lymphocytes were determined by histocompatibility typing (Columbia University, Histocompatibility Typing Laboratory, New York, NY). Cell populations Peripheral blood mononuclear cells were isolated from the leukapheresis specimens by Ficoll-hypaque flotation. Neoplastic T cells and CD8 T-cell lines were isolated and propagated using magnetic bead technology and cytokine stimulation, as previously described (Berger et al., 1996). The purity of the cell populations was confirmed by flow cytometry and the neoplastic T cells were generally 95% Vb8a1 and the CD8 populations were on average 84% pure. The long-term CD8 lines used in this report were the same cell lines that had been assessed previously for cytotoxicity and shown to be specifically cytolytic for autologous tumor cells (Berger et al., 1996). B lymphocytes were isolated from the peripheral blood of the patients using CD19-conjugated magnetic beads (Dynal, Lake Success, NY) and transformed by incubation with Epstein-Barr virus obtained from the supernatant of a productively infected marmoset B-cell line. The transformed B lymphoblasts were phenotyped with the B-cell-specific CD19 MAb and were found to Grant sponsor: NIH; Grant numbers: R01-CA43058 and P30-AR41942. *Correspondence to: Department of Dermatology, Yale University School of Medicine, 333 Cedar Street, New Haven, CT 06519-8059, USA. Fax: (203) 785-7637. E-mail: [email protected] Received 19 September 1997; Revised 15 December 1997 TUMOR-SPECIFIC PEPTIDES IN CTCL TABLE I ? AMINO ACID SEQUENCE OF THE Vb8 TCR Vb8 = IYFNNNVPIDDSGMPEDRFSAK Native peptide: MPNASFSTLKIQPSEPRD ;Vb8 0 ;D = 0 J2.1 = AR idiotypic peptide: SAVYFCASS LIGG SYNEQF ;Vb8 0 ;D= 0 J2.3 = SS idiotypic peptide: SAVYFCASS FV MRYTQY 0 C2= FGPGTRLTVLEDLKNVFPPEV1 1Conserved sequence is presented in medium type and idiotypic and native peptides are in bold. be .95% CD191. The B-cell lines were maintained in RPMI 1640/10% fetal calf serum (FCS; GIBCO, Gaithersburg, MD) and used to present peptides to CD8 T cells in the enzyme-linked immunosorbent assay (ELISA). Dendritic cells were isolated from the peripheral blood of the patients by Ficoll-hypaque flotation and incubated overnight with granulocyte-macrophage colony stimulating factor (GM-CSF; 0.05 ng/ml) and interleukin 4 (IL-4; 800 units/ml; R&D Systems, Minneapolis, MN), in RPMI 1640/10% FCS. The floating cells were removed and the adherent population was recultured in the same media containing cytokines and used as peptide-presenting cells after 8 days of in vitro culture. The dendritic cell lineage of the cultured cells was confirmed by phenotyping with MAbs (.90% reactive with MAbs identifying CD11c, class I and II, CD80; non-reactive with T- or B-cell-specific antibodies). Idiotypic peptides The sequence of the patient?s TCR was determined by cDNA sequencing, as previously described (Longley et al., 1995). Potential target CDR3 peptides were identified by examination of the sequence of the b chain of the patient?s TCR CDR3 for protein sequences that could be accommodated by the peptide-binding motifs of the class I molecules expressed on the patient?s cells. Motif-conforming CDR3 peptides were synthesized with a RANIN Symphony peptide synthesizer, purified by reversed-phase high performance liquid chromatography (RP-HPLC) and their purity confirmed by mass spectrometry in the Yale University Keck Protein Synthesis Facility. Evaluation of synthetic idiotypic peptides To determine if cytotoxic T cells recognized synthetic idiotypic peptides, 2 3 104 autologous B lymphoblasts were g-irradiated (2,000 rads, cesium source) and pulsed with 10 痞 of the optimal concentration of peptide (predetermined by a dose-response curve) in RPMI 1640/0.5% bovine serum albumin (BSA), containing 30 ng/ml b2-microglobulin, and incubated at 37蚓 for 1 hr. CD8 T-cell lines (predetermined to be cytolytic for autologous tumor cells; Berger et al., 1996) were added to the pulsed lymphoblasts at varying effector to target ratios. The cultures were incubated at 37蚓 for 18 hr, centrifuged and the supernatant fluid was harvested and tested for release of tumor necrosis factor-a (TNF-a) in a colorimetric ELISA according to the manufacturer?s directions (R&D Systems). The plates were analyzed in an ELISA reader (Sigma, St. Louis, MO). In class I blocking studies, the B-lymphoblast targets were preincubated, on ice, with a murine MAb W6/32 (1:10 dilution of culture supernatant, HB95; ATCC, Rockville, MD) or an isotype control antibody of irrelevant specificity. The cells were washed prior to peptide pulsing and the TNF-a ELISA was performed as described above. 305 Extraction of class I peptides from neoplastic CTCL T lymphocytes Approximately 100 3 106 CTCL tumor cells were centrifuged (900g), washed and the pellet was resuspended in 10 ml of 0.1% trifluoroacetic acid (TFA). The pelleted cells were ground in a tissue grinder and transferred to a Dounce homogenizer. After 20 passes of the homogenizer, the homogenate was transferred to a sonicator and sonicated twice with 10 sec bursts. The pH of the sonicate was adjusted to 2.0 by the addition of 1% TFA. The suspension was incubated on ice for 30 min with periodic manual agitation and then centrifuged (2,600g, 30 min, 4蚓) in an ultracentrifuge (Beckman, Palo Alto, CA) and the supernatant fluid lyophilized. The lyophilized supernatant was dissolved in 2 ml of 0.1% TFA and filtered through a Centricon-10 concentrator (,10 kDa filtrate; Amicon, Beverly, MA). The filtrate was fractionated on an RP-HPLC unit (Gilson, Middleton, WI) with a stepwise linear acetonitrile (Aldrich, Milwaukee, WI)-HCL gradient using a Waters (Milford, MA) Delta-PAK 3.9 3 150 mm C-18 column. Fractions (0.5 ml) were collected at 1 min intervals and the liquid evaporated in a SpeedVac (Savant, Farmingdale, NY). The fractions were dissolved in RPMI 1640/0.5% BSA for testing in a TNF-a ELISA. TNF-a ELISA of HPLC fractions To identify fractions containing peptides that were stimulatory for autologous CD8 lines, 2 3 104 autologous B lymphoblasts were g-irradiated (2,000 rads, cesium source), pulsed with 20 痞 of the HPLC fractions in the presence of b2-microglobulin and incubated at 37蚓 for 1 hr. The TNF-a ELISA conditions were identical to those previously described for testing the synthetic idiotypic peptides. Reactive fractions were refractionated by passage over a VYDAC C-18 2.1 3 250 mm column (Hesperia, CA), retested and consistently reactive subfractions were submitted for mass determination (Yale University, Keck Facility) by matrix-assisted laser desorption/ionization mass spectrometry (MALDI; Woods et al., 1995). Positive subfractions, containing peptides of a size appropriate for class I binding, were sequenced by Edman degradation. Identification and synthesis of native peptides To identify the protein of interest, the native peptide sequence was compared with known protein sequences (BLAST network service search) in the data bases (Brookhaven Protein Data Bank, Swiss Prot, PIR, Genpeptide, Kabat sequences of proteins of immunological interest and transcription factor protein data base). Since the highest degree of homology was obtained when the native amino acid sequence was compared to a sequence contained in the constant region of the Vb8 TCR, synthetic peptides were constructed based on the known TCR DNA sequence. The identified 7-mer native peptide sequence was extended by the addition of DNA-sequence predicted amino acids to provide the optimum length for class I binding and to add class I allele, motif-fitting anchor residues. RESULTS To identify immunogenic class I MHC-associated peptides on CTCL cells from the 2 patients, both theoretical and direct strategies were pursued. First, the tumor cell?s TCR Vb chains were sequenced and peptides derived from the idiotypic region were selected based on their predicted ability to provide anchor residues that would theoretically fit into the class I alleles expressed on the patient?s lymphocytes. The potential immunogenicity of these peptides was evaluated by pulsing the peptides onto autologous B lymphoblasts and testing their capacity to stimulate TNF-a release from cytotoxic CD8 T-cell lines that had previously been shown to be cytolytic for the patient?s tumor cells (Berger et al., 1996). Sequence of the idiotypic region of the Vb chain of the TCR in CTCL patients The sequence of the idiotypic region of the Vb chain of the TCR was determined in 2 CTCL patients. Studies of the neoplastic T BERGER ET AL. 306 cells from these 2 patients were facilitated by the large majority (83?92% Vb81 TCR; Berger et al., 1996) of lymphocytes being clonally expanded malignant cells. The conserved portion of the Vb chain was identical in both patients? TCR sequence, while a different diversity and joining region was encoded in the DNA of the two patients? cells (Table I). Synthesis of motif-fitting idiotypic peptides To identify the class I alleles expressed on the CTCL patient?s cells, the patient?s peripheral blood lymphocytes were histocompatibility typed (Table II). Only 4 of the 6 class I alleles (B7, B44, Cw4 and Cw7) expressed on patient SS?s cells had known peptidebinding motifs. Three candidate tumor-specific antigenic peptides were identified in the idiotypic CDR3 of the TCR b-chain sequence of patient SS?s malignant T cells. These 3 idiotypic peptides could be accommodated by the known peptide-binding motifs of SS?s Cw4 and Cw7 class I alleles. The 3 idiotypic peptide sequences synthesized for patient SS all contained a hydrophobic amino acid at position 9 (Table II). This is in accordance with a preference for a hydrophobic COOH anchor residue, which has been observed in most class I HLA alleles (Davenport and Hill, 1996). The 3 idiotypic peptides differed at positions 1?3 and these differences might influence optimal MHC-peptide amino-terminus interaction and thus the antigenicity of these peptides (Bednarek et al., 1991). Both Cw4 and Cw7 class I alleles have a preference for an aromatic amino acid at position 2, as an auxiliary anchor (Jung et al., 1993). The S2 and S3 peptides both contain aromatic amino acids at position 2 (S2: tyrosine, S3: phenylalanine) that might serve as secondary anchors. Position 6 in the binding groove of both Cw4 and Cw7 alleles has a preference for a hydrophobic amino acid and this preference is fulfilled only in the sequence of peptide S1 by the presence of an alanine residue. The internal sequence of amino acids was identical in all 3 peptides, but shifted in position. Therefore, differences in immunogenicity of these peptides might be attributed to changes in the anchoring positions that could result in reduced TCR access to potentially stimulatory internal amino acids and the shifted alignment of the internal amino acid sequence. Histocompatibility testing of patient AR?s lymphocytes revealed that 4 class I alleles (A11, A31, B52 and Cw6) had known peptide-binding motifs. Three segments of the CDR3 of the TCR b chain of CTCL cells from patient AR coded for peptides that were found to conform to the binding motif of his B52 or Cw6 class I alleles and were therefore selected as potential candidate antigenic peptides. The carboxy-terminal amino acids of 2 of the synthetic idiotypic peptides (R1 and R2) were preferred hydrophobic residues. Peptide R3 had a hydrophilic amino acid, glycine, at the carboxy terminus that might reduce the class I binding ability of this peptide. The B52 class I allele has a preference for an aromatic amino acid at position 3 (Rammensee et al., 1995) and 2 of the 3 idiotypic peptides fulfilled this criteria (R1: tyrosine, R2: phenylalanine). Peptide R1 differed from peptides R2 and R3 at the amino terminus due to the presence of an alanine residue and this difference might influence the binding and antigenicity of this peptide. The internal sequence of amino acids was identical in the 3 idiotypic peptides but shifted in position. Any observed differences in the antigenicity of these peptides might result from the shift in internal sequence of amino acids or the influence of differences at the amino or carboxy terminus on the ability of the peptides to bind to the class I allele. Control peptides were synthesized with appropriately positioned amino acids to serve as anchor residues that fit the class I alleles of the patients? cells. The intervening amino acid sequences, in the control peptides, theoretically would not be recognized by antiCTCL CD8 T cells. Control peptide C1 retained allele-binding anchor residues for the motifs of Cw4 and Cw7, but was considered unlikely to be recognized by anti-CTCL CD8 T cells, due to inversion of the internal peptide sequence. Control, peptide (C2), was a peptide isolated from human immunodeficiency virus type 1 (HIV-1) and was known to bind to the Cw4 allele (Johnson et al., 1993). Since this peptide was not derived from either patient?s TCR b chain, it was considered unlikely to be recognized by autologous CD8 T cells. Evaluation of the idiotypic peptides for cytotoxic T-cell recognition We had previously demonstrated that cytotoxic T-cell lines derived from these 2 patients were specifically cytolytic for autologous tumor cells and that this cytolysis was class I-restricted (Berger et al., 1996). We also demonstrated that the cytolytic CD8 T cells released TNF-a in response to their autologous tumor target (Berger et al., 1996) and that the level of cytokine released reflected the cytolytic capacity of the patient?s CD8 effector cells. The CD8 T-cell lines that demonstrated the highest percentage of cytolysis also released the greatest amount of TNF-a (Berger et al., 1996). Cytotoxicity assays, in which tumor cells are lysed by CD8 T cells, commonly yield low results unless the target cells are blasts. Because CTCL cells respond poorly to mitogens and do not become blasts, we chose to use the TNF-a release assay that we had shown to be as reliable as cytotoxicity in studying immunity in CTCL (Berger et al., 1996). These same CD8 T-cell lines have been in culture for more than 1 year and have subsequently lost cytolytic capacity but retained the ability to secrete TNF-a when stimulated with the appropriate tumor cells. Since we and others (Coulie et al., 1994) have shown that TNF-a release is a highly sensitive and reproducible assay that correlates well with cytotoxicity, we have evaluated the CD8 T-cell response to TCR-derived peptides with this assay system. The 3 candidate idiotypic antigenic peptides were synthesized based on the CDR3 sequence of patient SS?s Vb8 TCR (Table I) and pulsed onto patient SS?s autologous B-lymphoblast line. When a 10 然 concentration of each of the 3 idiotypic peptides was tested, only peptide S2 increased the production of TNF-a over the baseline response observed when CD8 cells were cocultivated with B cells without peptide (Fig. 1). This finding demonstrated that only peptide S2 was specifically recognized by the autologous CD8 T-cell lines. TABLE II ? TCR IDIOTYPIC PEPTIDES Patient HLA phenotype DNA predicted amino acid sequence SS A29/32, B7/44, Cw4/7 SAVYFCASSFV MRYT Vb8 D Jb2.3 AR A11/31, B52/57, Cw6/?6 SAVYFCASSLIGGSYNE Vb8 D1.1 Jb2.1 Idiotypic peptide sequence S1: AVYFCASSF(1,2,3)4 S2: VYFCASSFV1,2,3,4 S3: YFCASSFVM1,2(3,4) C1: VYSSFCAFV1,2(3)4 R1: AVYFCASSL(1,2)3,4 R2: VYFCASSLI1,2,3,4 R3: YFCASSLIG1,2,3 C2: SFNCGGEFF1,2(4)5 Peptide contains allele-specific sequence motif for: 1HLA-Cw4; 2HLA-Cw7; 3HLA-B52; and 4HLA-Cw6.? motif-fitting sequence (i.e., not all anchoring residues present).?5Known HLA-Cw4 binding T-cell epitope from HIV-1 gp120 (380?388) that can also fit the motif of the other alleles tested.?6? Second allele could not be identified with conventional typing sera. (1,2,3,4)Partial TUMOR-SPECIFIC PEPTIDES IN CTCL FIGURE 1 ? Identification of the S2 peptide as being immunogenic. Only the S2 idiotypic peptide stimulates the CD8 T-cell line S16, isolated from patient SS, to produce levels of TNF-a that exceed the background level found when CD8 cells are cultured with autologous B lymphoblasts without peptide. All idiotypic peptides were tested at a 10 然 concentration and the effector to target ratio (E/T) of CD8 cells to B lymphoblasts was 3:1. The CD8 response to the S2 peptide was shown to be dosedependent (Fig. 2), with a peptide concentration as low as 1 nM producing TNF-a levels that exceeded background release. The response to the S2 peptide was not restricted to a single CD8 line established from patient SS. Four independently isolated CD8 lines tested at different time points consistently released increased levels of TNF-a when stimulated with either the 10 然 or 1 nM concentration of the S2 idiotypic peptide loaded on autologous B cells (Figs. 1?3). Moreover, the response was shown to be specific for the S2 peptide, with significantly elevated TNF-a release demonstrated at both 10 然 and 1 nM concentrations of the S2 peptide in comparison to the levels found with 3 controls. The controls tested were a control peptide, C1, which was predicted to bind to patient SS?s class I alleles, but not stimulate due to an internally inverted sequence of amino acids; patient AR?s idiotypic peptide, R2, which was derived from his CDR3 region and differed in sequence from the S2 peptide at the 2 carboxy-terminal amino acids; and as a background control, CD8 cells admixed with B cells without peptide (Fig. 3). These results, viewed in the context of the non-recognition of the other S1 and S3 idiotypic peptides, indicate the high level of specificity of the CD8 response for the S2 idiotypic peptide. Three peptides derived from the idiotypic sequence of the Vb region of the TCR of patient AR (Table II) were also tested by serial dilution and pulsing onto autologous B lymphoblasts. One of the 3 peptides, R1, was identified as immunogenic by the R8 CD8 T-cell line (effector to target ratio of 21:1). The concentration of TNF-a released from AR?s CD8 T-cell line was significantly ( p # 0.001, 10 然, 1 nM; p # 0.01, 0.1 nM) increased when peptide R1 was added to autologous B lymphoblasts in comparison to the level found when a control motif-fitting peptide C2, or parallel concentrations of other idiotypic peptides that differed in sequence by only 2 amino acids (R2, R3; Table II) or CD8 T cells added to B lymphoblasts without peptide were tested. Class I blocking of idiotypic TCR peptides To confirm that the CD8 response elicited in the TNF-a release assay was related to class I binding of idiotypic TCR peptides, the 307 FIGURE 2 ? Dose-response curve of the S2 peptide. Concentrations of the idiotypic S2 peptide ranging from 10 然 to 1 nM stimulated the S5 CD8 T-cell line to release levels of TNF-a that exceed the background level found when CD8 cells were cultured with B lymphoblasts without peptide. The effector to target ratio (E/T) was 6:1. FIGURE 3 ? Specificity of the S2 peptide. The S2 peptide was tested at 10 然 and 1 nM concentrations with 2 independently isolated CD8 T-cell lines. The S2 peptide stimulated significantly more ( p # 0.03 and p # 0.004) TNF-a release from both CD8 lines than the control peptide (C1) or an idiotypic peptide synthesized from the CDR3 of patient AR (R2) or the baseline control of CD8 cells cultured with B lymphoblasts without peptide. The effector to target ratios (E/T) were 5:1 for the S20 CD8 line and 3:1 for the S11 CD8 cell line. capacity of the peptides to access the class I binding groove was blocked by preincubation of the B lymphoblasts with an antibody that recognizes all class I molecules (W6/32). Autologous B lymphoblasts from patient SS or AR were preincubated with an anti-class I antibody or an isotype-matched control murine ascites of irrelevant specificity. Recognition of peptide S2 by a CD8 T-cell line was reduced to control levels after the class I molecules on the autologous B lymphoblasts were blocked with the anti-class I 308 BERGER ET AL. FIGURE 4 ? Class I blocking of the response to the S2 idiotypic peptide. Preincubation of the B lymphoblasts with the W6/32 MAb inhibited the ability of peptide S2 to bind and stimulate release of TNF-a from the CD8 T-cell line S21. Incubation of the B cells with murine ascites (ASC) of the same isotype as W6/32 but with an irrelevant specificity does not inhibit the binding of the S2 peptide, resulting in elevated release of TNF-a above the background level found when CD8 cells are cultured with B cells without peptide or a control peptide (C1) or an idiotypic peptide synthesized from the CDR3 of the other CTCL patient, AR (R2). The effector to target ratio (E/T) was 5:1. FIGURE 5 ? Evaluation of HPLC fractionated TFA-eluted class I peptides. Class I peptides were eluted from CTCL patient AR?s tumor cells with TFA and fractionated with a RP-HPLC. The fractions were pulsed on autologous B lymphoblasts and tested with AR?s CD8 T-cell line R8. Several reactive fractions (25, 26 and 29) reproducibly stimulated increased release of TNF-a over that found when CD8 cells were cultured with B cells without peptide. The effector to target ratio (E/T) was 8:1. antibody, W6/32 (Fig. 4). Incubation of the B lymphoblasts with control murine ascites did not inhibit the binding or presentation of the S2 peptide and resulted in enhanced release of TNF-a. Class I blocking of the TNF-a response to the S2 peptide was confirmed with 2 additional independently isolated CD8 T-cell lines (S3 and S12; results not presented). In a similar fashion, the R8 CD8 T-cell response to the R1 peptide was inhibited by class I blocking with W6/32, while preincubation of the B lymphoblasts with murine ascites permitted enhanced TNF-a release over control levels (C2 peptide or CD8 cells cocultivated with B cells without peptide; results not presented). These results demonstrate that recognition of idiotypic TCR peptides is dependent on effective class I loading and recognition. Isolation and sequence determination of native CTCL peptides The presence of an immune response in CTCL patients to synthetic idiotypic TCR peptides suggests that the tumor cells might present native peptides that could be recognized as immunogenic by CD8 anti-tumor cell lines. To pursue this possibility, we isolated native peptides from class I molecules on the CTCL tumor cells and determined their sequence. TFA-extracted peptides from patient AR?s tumor cells were fractionated using a RP-HPLC and the fractions tested for CD8 recognition by loading on autologous B lymphoblasts and measuring TNF-a release. Several reactive fractions were identified (Fig. 5) and fractions 21?25 were found to contain peptides of an appropriate molecular mass by MALDI which indicated that they could have been derived from 8?11 amino acid class I peptides. These fractions were subfractionated, retested and several reactive subfractions identified (Fig. 6) that contained peptides that were identical in mass to the peptides found in the original fractions. Subfraction 25-3 contained sufficient material for sequencing and a 7 amino acid peptide sequence was identified (Table III). This sequence, MPRASES, matched 5 of 7 amino acids encoded in the FIGURE 6 ? Evaluation of subfractions of the original reactive fraction 25. Fraction 25 was subfractionated by RP-HPLC and several reproducibly reactive subfractions were identified (2, 3, 5 and 6). Subfraction 25-3 stimulated increased TNF-a release above that found when R8 CD8 cells were cultured with B cells without peptide at an effector to target ratio (E/T) of 8:1. Subfraction 25-3 was found to contain peptides that were identical in mass, by MALDI, to those found in the original fraction 25 and was submitted for sequencing. conserved portion of the TCR Vb8 chain. No known protein could be identified in the protein data base that provided a better match for this sequence, suggesting that the identified peptide antigen is novel and may be unique to this particular set of CTCL cells. Since the sequence lacked appropriate anchor residues and was shorter than the optimum class I groove-fitting length of 8?10 amino acids, additional residues based on the known DNA sequence for the Vb8 chain were added at the carboxy terminus. TUMOR-SPECIFIC PEPTIDES IN CTCL 309 TABLE III ? NATIVE PEPTIDE SEQUENCE Edman sequence: DNA sequence: Synthesized peptides V-1: V-2: V-3: 1Divergent MPRASES1 MPNASFSTLKI MPNASFSTLKI MPRASESLK MPRASESTL sequences are noted in bold. CD8 T-cell recognition of native peptides Three peptides were synthesized and tested to determine if they were recognized as immunogenic by AR?s CD8 T-cell lines. The synthesized V-1 peptide contains the entire 11 amino acid sequence that was predicted by the DNA-encoded TCR constant region sequence, with the amino acids asparagine and phenylalanine replacing the amino acids argenine and glutamic acid that were identified in the native peptide sequence. The carboxy-terminal amino acids leucine, lysine and isoleucine provide anchor residues that can be accommodated in AR?s A11, B52 or Cw6 class I alleles. Leucine and isoleucine are hydrophobic residues and are therefore preferred carboxy-terminal anchor residues. Lysine is a positively charged amino acid and could represent a preferred anchor for the motif of the class I A-11 allele (Falk et al., 1994). Although 11 amino acids exceeds the optimal 8?10 class I peptide size, the internal amino sequence might be accommodated, particularly in the A11 allele which is known to bind longer peptides, through internal bulging of the peptide chain, between the anchor residues. Peptides exceeding 11 amino acids in length have been isolated from class I molecules (Falk et al., 1994). The V-2 peptide contains the native peptide sequence with the addition of 2 motif-fitting anchor residues, leucine and lysine. The threonine residue was omitted to provide a length of 9 amino acids while preserving the availability of the 2 strong anchor residues for the A11, B52 or Cw6 class I alleles. Peptide V-3 extends the native peptide sequence with the next 2 DNA predicted amino acids threonine and leucine, providing one strong anchor residue for the B52 or Cw6 alleles. When these peptides were tested, the V-1 peptide was not recognized, while both V-2 and V-3 provoked the CD8 lines to release elevated amounts of TNF-a which were similar to the level of TNF-a produced by the initial peptide pool (70?80 pg/ml; Fig. 7). In addition, the V-2 and V-3 peptides stimulated increased release of TNF-a from AR?s CD8 T cells when compared to the levels obtained when the synthetic R1 peptide was tested (20?35 pg/ml). These results imply but do not conclusively prove that peptides partially derived from the conserved portion of the TCR in CTCL patients may serve as immunogens for autologous CD8 T cells and may be useful in dendritic cell-based vaccine protocols designed to boost the immune response to the tumor cells. Recognition of the native peptide sequence by CD8 cells from other Vb81 CTCL patients Since the native sequence isolated from tumor cells of patient AR was homologous to the conserved sequence of the Vb8 chain, this sequence was also present in the TCR of patient SS. We tested whether CD8 cells from patient SS could recognize the native peptides when they were presented by autologous dendritic antigenpresenting cells, a prerequisite for utility of this sequence as a vaccine component. A CD8 cell line from patient SS recognized peptide V-2 but not V-3 or a control peptide when the peptides were presented on autologous dendritic cells (Fig. 8). The V-2 peptide sequence contains at least 1 anchor residue consistent with motifs found in patient SS?s B7, Cw4 and Cw7 class I alleles. These results indicate that the native peptide is recognized by CD8 lines isolated from 2 different CTCL patients whose malignant T cells express a Vb8 TCR containing the conserved sequence from which the native peptide was synthesized. FIGURE 7 ? Evaluation of the synthesized native peptides. Three peptides were synthesized based on the native peptide sequence extended with DNA sequence-predicted amino acids. Peptides V-2 and V-3 pulsed on AR?s B lymphoblasts stimulated increased release of TNF-a from AR?s CD8 T-cell line R9 at an effector to target ratio (E/T) of 9:1. The amount of TNF-a produced by the CD8 line when it was presented with either the V-2 or V-3 peptide exceeded that produced by presentation of the V-1 peptide or when CD8 cells were cultured with B cells without peptide. FIGURE 8 ? Recognition of the native peptides pulsed on dendritic cells by SS?s CD8 cells. Native peptide V-2 isolated from patient AR?s tumor cells also stimulated a CD8 T-cell line S18 from patient SS to produce elevated levels of TNF-a in comparison to the level found with the V-3 peptide, a control peptide (C1) or CD8 cells cultured with dendritic cells without peptide. The effector to target ratio (E/T) was 2:1. The culture supernatants were diluted 1:2 to achieve values within the reference range of the standard curve of the assay, due to the high background level of TNF-a produced by the dendritic cells alone. DISCUSSION Our results reveal that synthetic peptides, designed to represent segments of the CDR3 idiotypic region of the b chain of clone-specific TCR of CTCL cells, are recognized by autologous 310 BERGER ET AL. CD8 cells as immunogenic. This finding supports the possibility that clonotypic TCR proteins are at least one source of tumorspecific antigenic peptides in CTCL. We also report that an antigenic native peptide obtained directly from CTCL cells is a novel protein, partially homologous in amino acid sequence to a segment of the variable region of the clonal TCR expressed on the same population of CTCL tumor cells. Although this peptide is not definitively derived from TCR proteins, this is direct evidence that CTCL cells display immunologically relevant peptides. The TCRs of malignant T cells provide a potential source of tumor antigens in T-cell malignancy, since these protein heterodimers are clonotypically expressed at high levels on the tumor cell surface and are likely to be abundantly represented in the endoplasmic reticulum pool of candidate class I binding peptides. Both murine and human cytotoxic CD81 T-cell clones can recognize peptides derived from the conserved region of the Vb of the TCR of autologous CD41 T cells (Jiang et al., 1995; Ware et al., 1995). Vaccination with peptides derived from the framework 3 region of the Vb8.2 TCR suppresses experimental autoimmune encephalomyelitis and collagen II-induced arthritis (Kumar et al., 1997). Extrapolation of these results to humans suggests that TCR-derived peptides may be clinically relevant targets for specific immunotherapy of both autoreactive and malignant T-cellmediated diseases. The TCR is a heterodimer consisting of an a and b chain (Rowen et al., 1996), made up of a constant and a variable region. The variable portions are composed of the complementarity determining regions that contact the MHC/peptide bimolecular complex. The third complementarity region (CDR3) encoded by the hypervariable region directly interacts with the MHC-held peptide antigen. Diversity in this region generates a unique patient-specific amino acid sequence that is clonotypically expressed by the TCR of the malignant T cells in CTCL. We have previously reported (Berger et al., 1996) that CD8 T cells can recognize antigenic peptides on autologous CTCL cells in the context of class I MHC glycoproteins. Our current results extend these observations and suggest that clonotypic peptide derivatives of the CTCL TCR are a source of such tumor-specific antigens. TCR peptides constructed under theoretical constraints imposed by MHC motif-binding requirements were able to sensitize autologous target B lymphoblasts, so that they could stimulate release of TNF-a from anti-CTCL CD8 T cells from the same individual. In some individuals with advanced CTCL, despite the expansion of malignant CD41 T cells, a functional cytotoxic CD8 T-cell compartment persists (Berger et al., 1996). Although, this CD8 cells are unable to eradicate the malignancy, they are specifically cytotoxic for the tumor cells and may play a role in ameliorating the course of the disease. Cytotoxic T lymphocytes recognize 8?10 amino acid peptides derived from proteasomal degradation of endogenous proteins that are transported into the endoplasmic reticulum by the TAP system, where they associate with class I MHC molecules, prior to export to the cell membrane. Class I MHC alleles have peptide-binding motifs that designate preferred positional anchor residues (Rammensee et al., 1995). If the amino acid sequence of the peptide contains appropriate amino acids that conform to the allele-specific motif, the peptide will fit in the class I binding groove. Based on pooled amino acid sequence data, it is possible to scan the sequence of potentially immunogenic peptides and select those that are most likely to be accommodated in a given class I allele (Rammensee et al., 1995). Binding of these candidate peptides to the class I molecules, and their capacity to sensitize target cells to attack by selective CD8 T cells, can then be assessed. Using this approach, we have demonstrated a high degree of specificity in the response to a sequence-predicted idiotypic peptide, as shown by the consistent recognition of a single peptide, by all the CD8 lines tested from an individual patient. The other candidate idiotypic peptides tested differed from immunogenic peptides by only 1?2 amino acids and were not effectively recognized by the CD8 cells. Substitutions at the amino or carboxy terminus of these peptides may have reduced their ability to bind to the class I allele or their stimulatory capacity for the autologous CD8 cells. These results suggest that immunization protocols should be tailored to optimally select reactive peptides and that simple conformation to MHC binding rules will not suffice for identification of relevant immunogenic epitopes. In our second strategy, a native peptide was identified by direct elution of class I molecules which bears a high degree of homology to the known sequence of the constant portion of the Vb8 TCR. The absence of total sequence homology suggests that this peptide might be derived from a novel protein that is not represented in the present data bases. Alternatively, the sequence divergence might arise from errors in the sequencing process due to the low concentration of the relevant peptide in the extracted pool. While the derivation of the native peptide from the TCR remains to be conclusively demonstrated, its recognition (after extension with Vb8 TCR DNA sequence-predicted amino acids) by tumorreactive CD8 T-cell lines suggests that this peptide may be a relevant epitope for therapy in CTCL. The specificity of the immune response in CTCL for the native peptide is supported by the restricted reactivity of the CD8 T-cell lines for CTCL tumor cells (Berger et al., 1996). The same CD8 T-cell lines used in this study were demonstrated to be cytotoxic for autologous tumor cells, and non-lytic when stimulated with normal lymphocytes, T-cell blasts or autologous B lymphoblasts. Therefore, the CD8 T-cell lines recognize peptides that are restricted to CTCL tumor cells. While a small normal Vb8 population in the peripheral blood of CTCL patients might be targeted by an anti-Vb tumor response, depletion of this subset would be acceptable in the context of reduction of the massively expanded tumor cell clone. The ability of dendritic cells to present one of the native peptides implies that peptide-pulsed dendritic cell vaccine immunotherapy may be useful in the treatment of CTCL. Because of their display of those accessory molecules which provide necessary second signals to initiate immune responses, dendritic antigen-presenting cells bearing tumor peptides in their class I molecules are efficient components of experimental anti-tumor cellular vaccines. To extend these findings, we are sequencing additional peptide pools derived from CTCL tumor cells, to identify additional candidate immunogenic peptides. In addition, murine studies in the 2B4.11 mouse model of T-cell malignancy are in progress to define the optimal approach for peptide vaccination against TCR epitopes. The existence of tumor-specific peptide antigens on malignant cells in CTCL, a relatively common T-cell malignancy, may permit the development of peptide-based immunotherapy for this disorder. ACKNOWLEDGEMENTS This work was supported by NIH grants R01-CA43058 and P30-AR41942. REFERENCES BAGOT, M., WECHSLER, J., LECS, M.C., REVUZ, J., FARCET, J.P. and GAULARD, P., Intra-epidermal localization of the clone in cutaneous T-cell lymphoma. J. Amer. Acad. Dermatol., 27, 589?593 (1992). BEDNAREK, M.A., SAUMA, S.Y., GAMMON, M.C., PORTER, G., TAMHANKAR, S., WILLIAMSON, A.R. and ZWEERINK, H.J., The minimum peptide-epitope from the influenza virus matrix protein: extra and intracellular loading of HLA-A2. J. Immunol., 147, 4047?4053 (1991). BERGER, C.L., WANG, N., CHRISTENSEN, I., LONGLEY, J., HEALD, P. and TUMOR-SPECIFIC PEPTIDES IN CTCL EDELSON, R.L., The immune response to class I associated tumor-specific cutaneous T cell lymphoma antigens. J. invest. Dermatol., 107, 392?397 (1996). COULIE, P.G., BRICHARD, V., VAN PEL, A., WO?LFEL, T., SCHNEIDER, J., TRAVERSARI, C., MATTEI, S., DE PLAEN, E., LURQUIN, C., SZIKORA, J.-P., RENAULD, J.-C. and BOON, T., A new gene coding for a differentiation antigen recognized by autologous cytolytic T lymphocytes on HLA-A2 melanomas. J. exp. Med., 180, 35?42 (1994). DAVENPORT, M.P. and HILL, A.V.S., Peptides associated with MHC class I and class II molecules. In: M. Browning and A. McMichael (eds.), HLA and MHC: genes, molecules and function, pp. 277?308, Bios Scientific, Oxford (1996). EDELSON, R.L., and 16 OTHERS, Treatment of leukemic cutaneous T cell lymphoma with extracorporeally-photoactivated 8-methoxypsoralen. N. Engl. J. Med., 316, 297?303 (1987). EDELSON, R.L., BERGER, C.L., RAAFAT, J. and WARBURTON, D., Karyotype studies of cutaneous T cell lymphoma. Evidence for clonal origin. J. invest. Dermatol., 73, 548?550 (1979). EDELSON, R.L., HEALD, P., PEREZ, M. and BERGER, C., Extracorporeal photochemotherapy. Biol. Ther. Cancer, 4, 1?12 (1994). FALK, K., ROTZSCHKE, O., GRAHOVAC, B., SCHENDEL, D., STEVANOVIC, S., GNAU, V., STEVANOVIC, S., JUNG, G. and RAMMENSEE, H.-G., Peptide motifs of HLA-A1, -A11, -A31, and -A33 molecules. Immunogenetics, 40, 238?241 (1994). HEALD, P., ROOK, A., PEREZ, M., KNOBLER, R., JEGASOTHY, B., GASPARRO, F., BERGER, C. and EDELSON, R., Treatment of erythrodermic cutaneous T cell lymphoma with extracorporeal photochemotherapy. J. Amer. Acad. Dermatol., 27, 427?433 (1992). JIANG, H., WARE, R., STALL, A., FLAHERTY, L., CHESS, L. and PERNIS, B., Murine CD81 T cells that specifically delete autologous CD41 T cells expressing Vb8 TCR: a role of the Qa-1 molecule. Immunity, 2, 185?194 (1995). JOHNSON, R.P., TROCHA, A., BUCHANAN, T.M. and WALKER, B.D., Recognition of a highly conserved region of human immunodeficiency virus type 1 gp 120 by an HLA-Cw4-restricted cytotoxic T-lymphocyte clone. J. Virol., 67, 438?445 (1993). JUNG, G., STROMINGER, J.L. and RAMMENSEE, H.-G., Allele-specific peptide 311 ligand motifs of HLA-C molecules. Proc. nat. Acad. Sci. (Wash.), 90, 12005?12009 (1993). KUMAR, V., AZIZ, F., SERCARZ, E. and MILLER, A., Regulatory T cells specific for the same framework 3 region of the Vb8.2 chain are involved in the control of collagen II-induced arthritis and experimental autoimmune encephalomyelitis. J. exp. Med., 185, 1725?1733 (1997). KUNG, P.C., BERGER, C.L., GOLDSTEIN, G., LOGERFO, P. and EDELSON, R.L., Cutaneous T-cell lymphoma: characterization by monoclonal antibodies. Blood, 57, 261?266 (1981). LONGLEY, J., TYRELL, L., LU, S.-Z., FARRELL, J., DING, T.-G., YAN, S., SALLEE, D., HEALD, P., BERGER, C., TIGELAAR, R. and EDELSON, R., Malignant and normal T cells show random use of T cell receptor a chain variable regions in patients with cutaneous T-cell lymphoma. J. invest. Dermatol., 105, 62?64 (1995). PICKER, L.J., MICHIE, S.A., ROTT, L.S. and BUTCHER, E.C., A unique phenotype of skin-associated lymphocytes in humans: preferential expression of the HECA-452 epitope by benign and malignant T cells at cutaneous sites. Amer. J. Pathol., 136, 1053?1068 (1990). RAMMENSEE, H.-G., FRIEDE, T. and STEVANOVIC, S., MHC ligands and peptide motifs: first listing. Immunogenetics, 41, 178?228 (1995). ROWEN, L., KOOP, B.F. and HOOD, I., The complete 685-kilobase DNA sequence of the human b T cell receptor locus. Science, 272, 1755?1762 (1996). WARE, R., JIANG, H., BRAUNSTEIN, N., KENT, J., WIENER, E., PERNIS, B. and CHESS, L., Human CD81 T lymphocyte clones specific for T cell receptor Vb families expressed on autologous CD41 T cells. Immunity, 2, 177?184 (1995). WOOD, G.S., DUBIEL, C., MUELLER, C., ABEL, E.A., HOPPE, R.T., EDINGER, A., WEISSMAN, I. and WARNKE, R.A., Most CD81 cells in skin lesions of CD31 CD41 mycosis fungoides are CD31 T cells that lack CD11b, CD16, CD56, CD57, and human Hanukah factor mRNA. Amer. J. Pathol., 138, 1545?1552 (1991). WOODS, A.S., HUANG, A.Y.C., COTTER, R.J., PASTERNACK, G.R., PARDOLL, D.M. and JAFFEE, E.M., Simplified high-sensitivity sequencing of a major histocompatibility complex class I-associated immunoreactive peptide using matrix-assisted laser desorption/ionization mass spectrometry. Anal. Biochem., 226, 15?25 (1995).