Identification of novel CTL epitopes of CMV-pp65 presented by a variety of HLA alleles

Late-onset cytomegalovirus (CMV) disease in hematopoietic stem cell transplant recipients (later than day 100 after transplantation) remains a major cause of morbidity and mortality, despite the introduction of new antiviral agents; indeed, recent reports have rather indicated an increase in disease.^1-5  Absence of reconstitution of CMV-specific T-cell response at 3 to 4 months after transplantation or the use of immunosuppressive drugs is strongly associated with reactivation of CMV and the subsequent CMV disease.^2-9  Immunologic treatments, such as adoptive transfer of CMV-specific cytotoxic T-lymphocyte (CTL) clones^6,10  or CMV-specific T-cell lines,¹¹  have successfully protected patients at risk from CMV disease, indicating that T-cell immunity plays an important role in controlling CMV infection. Thus, immunologic monitoring of T-cell immunity against CMV is crucial to evaluate the status of immunocompromised patients.

Identification of the CTL epitopes derived from CMV is very valuable not only for monitoring antiviral immunity but also for ex vivo generation of antiviral CTLs for possible application in adoptive immunotherapy. Immunodominance of pp65 protein among CMV antigens has been reported,^12-14  but previously identified CTL epitopes derived from the pp65 protein were limited to frequent human leukocyte antigen (HLA) class I alleles. Moreover, the pp65 CTL epitope presented by HLA-A*2402, which is the most frequent allele in Japanese individuals, may not be immunodominant because the percentage of CD8⁺ T cells detectable with the HLA-A*2402/pp65 tetramer in healthy seropositive individuals is relatively low (A*2402, 0.1%¹⁵ ; versus A*0201, 0.75%; and B*0702, 1.85%¹⁶ ), whereas A*0201- or B*0702-restricted pp65 epitopes are considered immunodominant in white individuals. Therefore, additional pp65 epitopes of clinical significance need to be identified.

We previously reported an efficient strategy for in vitro CTL generation starting with as little as 10 mL of blood.¹⁷  By use of retroviral transduced CD40-activated B (CD40-B) cells as antigen-presenting cells (APCs), pp65-specific CTL lines could be generated from all of 4 CMV-seropositive healthy donors and found to be restricted by multiple HLA class I alleles, suggesting utility for epitope identification. In the present study, using a total of 11 pp65-specific CTL lines, including 7 newly generated ones, we attempted to identify novel CTL epitopes by enzyme-linked immunospot (ELISPOT) assay using stimulator cells transfected with truncated forms of the pp65 gene or linear expression fragments encoding various regions of the gene, with or without the help of computer algorithm–based epitope prediction. This approach was sufficiently sensitive to identify even the subdominant epitopes recognized by the CTL lines. Through this study, a total of 14 novel CTL epitopes were identified. Immunogenicity of a part of the newly identified epitopes was validated by successful generation of pp65-specific CTL lines from CMV-seropositive and also from some CMV-seronegative donors.

Peripheral blood samples were obtained from 11 CMV-seropositive and 8 CMV-seronegative healthy donors after we obtained informed consent. The study was approved by the institutional review board of the Aichi Cancer Center. Informed consent was provided according to the Declaration of Helsinki. CMV seropositivity was analyzed with regard to the presence of CMV-specific immunoglobulin G (IgG) using an enzyme-linked immunosorbent assay, and HLA typing was carried out at The HLA Laboratory (Kyoto, Japan; Table 1). Peripheral blood mononuclear cells (PBMCs) were isolated from peripheral blood by centrifugation on a Ficoll (Amersham Biosciences, Uppsala, Sweden) density gradient, and CD8⁺ and CD8^– fractions were separated using CD8 MicroBeads (Miltenyi Biotec, Bergisch Gladbach, Germany) and cryopreserved until use.

Plasmids, pcDNA3-pp65, pcDNA3–enhanced green fluorescent protein (EGFP), and pcDNA3.1 (Invitrogen, Tokyo, Japan) encoding HLA class I cDNA were constructed as previously described.^17,18  To generate pEAK10-pp65, a portion of pp65 DNA was transferred from pcDNA3-pp65 into the pEAK10 vector (Edge Biosystems, Gaithersburg, MD). A plasmid containing a mutant pp65 gene defective for the immunodominant HLA-A*0201–restricted epitope (NLVPMVATV to NLVPMVAAATV) was constructed by polymerase chain reaction (PCR)–based mutagenesis using the following primers: sense, ATGGTGGCTGCAGCTACGGTTCAGGGTCAG; antisense, AACCGTAGCTGCAGCCACCATGGGCACCAG (the inserted nucleotides are underlined). This plasmid was termed pcDNA3-pp65^ΔNLV. All peptides were purchased from Toray Research Center (Tokyo, Japan).

CD40-B cells were generated as previously described.^17,19  In brief, a thawed CD8^– fraction of PBMCs was cultured on a γ-irradiated (96 Gy) human CD40L-transfected NIH3T3 cell line²⁰  (t-CD40L; kindly provided by Dr Gordon Freeman, Dana-Farber Cancer Institute, Boston, MA) in the presence of interleukin 4 (IL-4; 4 ng/mL, Ono Pharmaceutical, Osaka, Japan) and cyclosporin A (CsA; 0.7 μg/mL, Sandoz, Basel, Switzerland) in 2 mL of Iscove modified Dulbecco medium (Invitrogen) supplemented with 10% pooled human serum. The expanding cells were transferred onto freshly prepared t-CD40L cells and fed with cytokine-replenished medium without CsA every 3 to 4 days. LCLs were established from the CD40-B cells with supernatant of an Epstein-Barr virus (EBV)–producing cell line (B95-8; American Type Culture Collection, Manassas, VA) in RPMI 1640 (Invitrogen) supplemented with 10% fetal calf serum (FCS; IBL, Takasaki, Japan), referred to as “complete medium.”

Retroviral transduction was conducted as previously described.¹⁷  In brief, the retroviral construct, LZRSpBMN/pp65 (the backbone plasmid, LZRSpBMN-Z was kindly provided by Dr G. Nolan, Stanford University, Stanford, CA), pLBPC/pp65, or pLBPC/EGFP, was packaged in the Phoenix gibbon ape leukemia virus (GALV) cell line²¹  (a gift from H.-P. Kiem, Fred Hutchinson Cancer Research Center; and from G. Nolan, Stanford University, Stanford, CA) using FuGENE 6 (Roche Diagnostics, Mannheim, Germany). CD40-B cells and LCLs were infected with the retroviral supernatant in the presence of 10 μg/mL polybrene (Sigma, Chicago, IL), spun at 1000g for 1 hour at 32°C, and incubated. Two days after, LCLs transduced with pp65 (LCL/pp65) or EGFP (LCL/EGFP) were selected in the presence of puromycin (0.7 μg/mL; Edge Biosystems). Transduction efficiency were assessed as previously described.¹⁷ 

Thawed CD8⁺ cells (1 × 10⁶) were cocultured with γ-irradiated (33 Gy) autologous pp65-transduced CD40-B (CD40-B/pp65) cells (1 × 10⁶) in 2 mL RPMI 1640 supplemented with 10% pooled human serum and recombinant human IL-7 (50 U/mL; Genzyme, Cambridge, MA) at 37°Cin 5% CO₂. On days 7 and 14, CD8⁺ cells were restimulated and one day after each stimulation recombinant human IL-2 (Chiron, Emeryville, CA) was added to the cultures at the final concentration of 20 U/mL. If necessary, rapidly growing cells were split into 2 to 3 wells and fed with fresh media containing 20 U/mL IL-2. Peptide-pulsed CD40-B cells were also prepared by incubation with 10 μM peptides derived from pp65 and used as APCs. For generation of CTL lines from seronegative donors, IL-12 (5 ng/mL; R&D systems, Minneapolis, MN) was added on day 0.

To construct deletion mutants of the pp65 gene, the plasmid, pcDNA3-pp65, was opened with ApaI and BamHI, and progressive 3′ deletions were produced by exonuclease III treatment using the Erase-a-Base System (Promega, Madison, WI). After ligation each clone was sequenced. A total of 22 clones were selected and termed “Δpp65(1-XXX)”; XXX indicates the amino acid position of the C-terminus of each clone.

The epitopes within the defined regions of the pp65 protein from human CMV strain AD169 were predicted by “HLA Peptide Binding Predictions” on the BioInformatics & Molecular Analysis Section (BIMAS) website^22,23  and also by “HLA Epitope binding prediction” (beta testing version) in the HLA Ligand/Motif Database.²⁴ 

Linear expression fragments encoding various regions of the pp65 gene were constructed by 2-step overlapping PCR. Targeted region-specific 5′ and 3′ primers incorporating additional sequences (single- and double-underlined, see below) were designed (eg, 5′primer, TCGGATCCACCATGCAGTACGATCCCGTGG [30 bp] and 3′ primer, GACTCGAGCGCTAGAAGAGCGCAGCCACGG [30 bp] for QYDPVAALF [amino acids (aa's) 341-349]) and used for PCR (KOD Plus; Toyobo, Osaka, Japan) with a template plasmid, pEAK10-pp65. CMV promoter (P_CMV) and bovine growth hormone (BGH) polyadenylation signal (pA) were independently amplified from pcDNA3.1 by PCR using the following primers: 5′ P_CMV, CTTAGGGTTAGGCGTTTTGC; 3′ P_CMV, NNCATGGTGGATCCGAGCTCGGTA; 5′ pA, NNTAGCGCTCGAGTCTAGAGGG; 3′ pA, GGTTCTTTCCGCCTCAGAAG; “N” means a mixture ofA/C/T/G. The 3′ P_CMV and 5′ pA primers contained overlapping sequences (underlined) with 5′ primer and 3′ primer, respectively, of the targeted region. The 3 PCR products, P_CMV, the targeted region, and pA, were conjugated by second PCR using primers, 5′ P_CMV and 3′ pA, and termed “P_CMVXXX-XXX” (each “XXX” indicates the amino acid positions of the N- or C-terminus).

ELISPOT assays were performed as described earlier.^17,25  In brief, a MultiScreen-HA plate (MAHA S4510; Millipore, Bedford, MA) was coated with antihuman interferon γ (IFN-γ) monoclonal antibody (M700A; Endogen, Woburn, MA) and used as an ELISPOT plate. The 293T cells were cotransfected with plasmids encoding each of the individual donor's HLA class I alleles and either pcDNA3-pp65, deletion mutants, or PCR products of the linear expression fragment by TransIT-293 (Mirus, Madison, WI) and were used as stimulator cells after 2 days. The transfected 293T, LCL/pp65, or LCL/EGFP cells (with or without peptide pulsing) were mixed with 10³ or more effector cells from the CTL lines generated. After cells had been incubated in 200 μL complete medium in a 96-well plate (3790; Costar Corning, Cambridge, MA) for 4 hours, all the aliquots were transferred into an ELISPOT plate and incubated for an additional 16 hours. To visualize spots, a biotin-labeled antihuman IFN-γ antibody (M701B; 1 μg/mL, Endogen), streptavidin-alkaline phosphatase (Biosource International, Camarillo, CA), and substrate were used. Spots were counted after computerized visualization using a scanner (Canon, Tokyo, Japan). For peptide titration assays, autologous LCL/EGFP cells were pulsed with various concentrations of synthetic peptides and then used as stimulator cells to see the differences in avidity of the effector cells from CMV-seronegative and CMV-seropositive donors.

LCL/pp65 or LCL/EGFP cells were labeled in 100 μL complete media with 3.7 MBq ⁵¹Cr for 1 hour at 37°C. Dermal fibroblasts were infected overnight with CMV (strain AD169) supernatant in the presence of 3.7 MBq ⁵¹Cr following 500 U/mL IFN-γ pretreatment for 24 hours. For peptide reconstitution assays, 1 μM of synthetic peptide was added 1 hour before introducing effector cells. After 4 hours incubation with effector cells, supernatants were counted in a gamma counter. The percentage of specific lysis was calculated as follows: [(experimental release – spontaneous release)/(maximum release – spontaneous release)] × 100.

We previously reported successful generation of pp65-specific CTL lines from all enrolled CMV-seropositive donors using retrovirally transduced CD40-B cells as APCs. In this study, we extended the findings by including additional 7 seropositive donors (Table 1). All CTL lines, with CD8⁺ phenotype, lysed pp65-transduced autologous LCLs efficiently, but not untransduced autologous LCLs, and the activities were inhibited by anti–HLA class I antibody (data not shown).

To determine the HLA restriction of these CTL lines we conducted ELISPOT assays using 293T cells transfected with the pp65 gene plus one of the HLA class I alleles belonging to donors as stimulator cells (Figure 1). For instance, a major population of the CTL line generated from donor P01 was stimulated by 293T cells transfected with HLA-B*1501 and the pp65 gene, while minor populations were restricted by HLA-A*1101 or -Cw*0401. Responses associated with the other HLA class I alleles were not detected when 10³ cells of the CTL line were used. The results demonstrated that the P01 CTL lines recognized multiple pp65-derived epitopes presented by at least 3 different HLA class I alleles. Similarly in the cases of P02, P03, P05, P08, and P11, the CTL lines were restricted by multiple HLA class I alleles and the sum of the spots restricted to each allele was comparable to those produced with autologous LCL/pp65. In contrast, in the cases of P04, P09, and P10, the sum of the spots restricted to each HLA class I allele was much smaller than those with pp65-transduced LCLs (Figure 1).

The CTL lines from HLA-A*0201–positive donors (P06, P07) well recognized the transfectant with pp65 alone (Figure 1 ▧) because 293T cells express HLA-A*0201 endogenously (data not shown). The mutant pp65 cDNA expression plasmid (pcDNA3-pp65^ΔNLV), lacking the HLA-A*0201–restricted immunodominant epitope, enabled us to detect CTL responses associated with HLA alleles other than HLA-A*0201. The results revealed that HLA-A*2402 and -B*4006 were subdominant restricting HLA alleles for P06 and P07 CTL lines, respectively (Figure 1 ▧).

As an initial step toward defining the epitopes recognized by these pp65-specific CTL lines, C-terminus–truncated pp65 genes were generated using the conventional exonuclease III deletion method to locate the regions encoding the epitopes (Figure 2). Thereafter, 293T cells were transfected with each of the deletion mutants plus restricting HLA class I cDNA and used for stimulation of each CTL line.

The P11 CTL line could recognize 293T cells transfected with HLA-A*2402 cDNA plus Δpp65(1-355) or longer deletion mutants, but not Δpp65(1-335) or shorter deletion mutants (Figure 2), indicating that the pp65 epitope presented by HLA-A*2402 should fully or partially be contained inside the region spanning between amino acid residues aa336 and aa355. Indeed, an HLA-A*2402–restricted epitope QYDPVAALF (aa's 341-349)¹⁵  is found within the region. The results for HLA-B*3501 and -B*5201 were also consistent with the reported epitopes (IPSINVHHY [aa's 123-131]²⁶  and QMWQARLTV [aa's 155-163],²⁷  respectively). These observations indicate that our strategy using deletion mutants worked effectively.

Next, we attempted to locate the regions containing pp65 CTL epitopes presented by other HLA class I alleles. As shown in Figure 2, such regions existed between the following amino acid residues: 514-561 (A*0207), 483-513 (A*1101), 212-234 (B*1501), 272-287(B*4002), 514-561 (B*4006), 356-378 (B*4403), 514-561 (B*5101), 1-40 (Cw*0102), 336-355 (Cw*0401), 187-211 (Cw*0801), 288-310 (Cw*1202), 187-211(Cw*1502). In the case of HLA-B*4001, the P03 CTL line strongly recognized not only Δpp65(1-287) or longer deletion mutants but also Δpp65(1-249), Δpp65(1-260), and Δpp65(1-271) to a lesser extent, suggesting the presence of 2 different HLA-B*4001–restricted epitopes. The existence of an additional subdominant epitope was also suggested in the cases of HLA-A*1101, -B*4403, -B*5101, and -Cw*0801 (Figure 2). As for HLA-Cw*1202, it is of note that the transfectants with the full-length plasmid produced a smaller number of spots than those with the shorter deletion mutants, such as Δpp65(1-482).

To predict the epitopes within the regions narrowed down by the deletion mutant experiments (Figure 2), amino acid sequences of the determined regions with a 10-aa extension to the N-terminus were analyzed by online computer algorithm software. The prediction results are listed in Table 2. For the HLA-A*0207–restricted epitope, A*0201 was alternatively selected because A*0207 was not available on the computer algorithm we used.^23,24  As for HLA-B*1501, the computer algorithm on the BIMAS website^22,23  depicted 3 candidate epitopes with similar scores (6, 4.4, and 4), one of which was also depicted by another algorithm²⁴  and subsequently adopted.

To test the recognition of the predicted epitopes by each CTL line, linear expression fragments encoding various peptides including the predicted epitopes were generated by overlapping PCR (see “Materials and methods”), transfected into 293T cells together with restricting HLA class I cDNA, and evaluated by ELISPOT assay. As shown in Figure 3, all predicted epitopes except for the HLA-B*5101–restricted one were well recognized by the corresponding pp65-specific CTL, and the specificity was confirmed using irrelevant fragments. For instance, the P01 CTL line could be stimulated by 293T cells cotransfected with a fragment encoding aa's 215 to 223 and HLA-B*1501 comparably to the case with the full-length pp65 gene, but not at all with aa's 227 to 235, indicating that aa's 215 to 223 (KMQVIGDQY; nonamer) is at least one of the HLA-B*1501–restricted epitopes. With HLA-B*4001, both of the 2 fragments encoding the predicted epitopes (aa's 232-240 and aa's 267-275) were well recognized by P03 CTL line (Figure 3), confirming the results of ELISPOT assay using deletion mutants (Figure 2).

In the case of HLA-B*5101, the fragments encoding the predicted epitope (aa's 547-555) or the one with the second highest score (aa's 545-553) were not or only poorly recognized by the P05 CTL line (Figure 3). Thus various other fragments were tested and the one encoding octamer peptide (DALPGPCI; aa's 545-552) was found to be well recognized. This octamer has a binding motif consistent with that for HLA-B*5101.²⁸ 

In summary, we successfully identified 11 new pp65-derived epitopes presented by 10 distinct HLA alleles (Table 3). Of interest is the fact that QYDPVAALF (aa's 341-349) restricted by HLA-Cw*0401 is identical to that restricted by HLA-A*2402.¹⁵ 

Since algorithms that predict peptides binding to HLA-Cw*0801, -Cw*1202, and -Cw*1502 are currently not available, we performed step-by-step epitope mapping using the linear expression fragments. Based on the results with deletion mutants and HLA-Cw*1202 (Figure 2), fragments encoding the region from aa 267 to aa 292 or to aa 302 were tested. The P04 CTL line recognized the transfectant with the fragment encoding aa's 267 to 302, but not that with aa's 267 to 292 (Figure 4A left), suggesting that the epitope should be fully or partially contained between aa 293 and aa 302. Next, various fragments encoding the septamer to dodecamer peptides within this region were generated and tested. The fragments whose C-terminus was Phe302 were well recognized, but those with His301 or Gly303 at the C-terminus were not or only poorly recognized (Figure 4C left). The CTL lines were partially activated by the fragments ending at Leu304. Progressive N-terminal deletion revealed that Ala295 was crucial for recognition. The same results were obtained with the P09 CTL line (data not shown). Additional experiments with synthetic peptides revealed that the CTL clone could recognize VAFTSHEHF (aa's 294-302) at 2-log lower concentrations than AFTSHEHF (aa's 295-302) (data not shown), indicating that the nonamer VAFTSHEHF is the minimal epitope presented by HLA-Cw*1202.

In the case of HLA-Cw*0801, fragments encoding the region up to aa 193, aa 204, or aa 208 were also constructed for further narrowing the regions identified (Figures 2 and 4). Because the fragment encoding aa's 173 to 208, but not aa's 173 to 204, was recognized by the P08 CTL line, the C-terminus of the epitope should be located between aa's 205 and 208 (Figure 4A middle). ELISPOT assay using transfectants with various linear expression fragments revealed that those encoding the peptides ending at Cys206 or Met208 were well recognized (Figure 4C middle). The results also indicated that Val198 was required for recognition. Thus we determined that VVCAHELVC (aa's 198-206) is the minimal epitope presented by HLA-Cw*0801. Interestingly, the same nonamer (VVCAHELVC) was independently identified as a minimal epitope presented by HLA-Cw*1502 (Figure 4C right).

To confirm the immunogenicity of the newly identified epitopes, we tried to generate pp65-specific CTL using synthetic peptide-pulsed CD40-B cells as APCs. After the third stimulation cytolytic activity of the CTL lines was assessed. A CTL line from donor P03, induced by CEDVPSGKL presented by HLA-B*4001, could lyse HLA-B*4001–positive allogeneic LCLs (from donor N06) expressing pp65 antigen and also the peptide-pulsed LCL/EGFPs (Figure 5A). Similar results were also obtained with the other 5 epitopes.

In the current and previous studies we failed to generate pp65-specific CTL lines from all 8 seronegative donors using pp65-transduced CD40-B cells as APCs. To improve stimulation efficiency, we applied CD40-B cells pulsed with a mixture of newly or previously identified epitopes. One million CD8⁺ T cells were stimulated 3 times with autologous CD40-B cells pulsed with a mixture of 2 to 4 synthetic peptides, which are presentable by HLA alleles belonging to each donor. The T-cell line from N02 produced IFN-γ spots after stimulation with 2 of the 4 peptides (Figure 6A). Peptide titration assay revealed that half-maximal number of spots for N02 line was obtained with approximately 3.0 nM of peptide IPSINVHHY, while that for P05 CTL line from a seropositive donor sharing HLA-B*3501 was 80 pM (Figure 6B top), indicating that a 38-fold higher peptide concentration was required in the N02 line. Nevertheless, the N02 could lyse autologous LCL/pp65s and also dermal fibroblasts infected with CMV although a lesser extent (9% at the effector-target [E/T] ratio of 25:1) (Figure 6C top). Two of the 4 peptides were recognized by the N03 T-cell line whose half-maximal spots for both peptides were obtained with 2.9 nM (peptide QYDPVAALF) and 14 nM (peptide HERNGFTVL), which were 35- and 77-fold higher than those for P02 CTL line from a seropositive donor sharing HLA-A*2402 and -B*4002, respectively. This N03 line lysed autologous LCL/pp65s and peptide-pulsed fibroblasts, but failed to lyse CMV-infected fibroblasts, suggesting the avidity of the line was not high enough for exerting cytotoxic activity against CMV-infected fibroblasts. On the other hand, in the case of N06 and N08, exogenously pulsed peptides (N06, CEDVPSGKL; N08, IPSINVHHY) on LCL/EGFPs were recognized (Figure 6A), whereas endogenously expressed pp65 in LCLs was not recognized (Figure 6A,C), which might partly be explained by the 10- to 20-fold higher peptide concentrations necessary for the half-maximal spots (65 nM for CEDVPSGKL in N06 and 36 nM for IPSNVHHY in N08; Figure 6B).

The present study focused on the systematic identification of novel CTL epitopes derived from CMV-pp65. We combined biotechnology with 22 C-terminal truncations of the pp65 gene at an average 22–amino acid interval (range, 11-48 aa's) and computer technology with algorithm-based prediction of peptide binding to certain HLA alleles. We were able to identify CTL epitopes using linear expression fragments constructed by overlapping PCR, even if a computer algorithm was not available for HLA alleles of interest. In general, to identify CTL epitopes, establishment of CTL clones and verification using ⁵¹Cr release assay have frequently been used. In this study, for efficient detection of CTL responses, we adopted the ELISPOT assay¹⁵  so that even responses to subdominant epitopes could be detected, such as the HLA-Cw*0102–restricted one in donor P03.

In the cases of P04, P09, and P10, the sum of the spots against 293T cells transfected with pp65 gene plus each HLA class I allele was unexpectedly smaller than that against pp65-transduced LCLs (Figure 1). This could be due to differential processing and/or presentation of pp65 proteins in LCLs and 293T cells. In this sense, HLA-A*2402 and -Cw*1202 are the focus as presenting molecules because both are present in the cases. Since the CTL lines were poorly stained with HLA-A*2402 tetramers incorporating peptide, QYDPVAALF (data not shown), it is unlikely that HLA-A*2402–restricted epitopes are concerned in this issue. Indeed, defective processing and/or presentation of the HLA-Cw*1202–restricted epitope from full-length pp65 in 293T cells was demonstrated (Figure 2). Probably, in LCLs, the pp65 is more efficiently processed and presented to yield the HLA-Cw*1202–restricted epitope. The better processing might attribute to so-called immunoproteasomes equipped by LCLs. So far, simultaneous transfection with immunoproteasome components, such as large multifunctional protease 2 (LMP2), LMP7, and LMP10, or IFN-γ treatment did not improve the recognition of 293T transfectants (data not shown), suggesting that additional factors are involved. The reason why truncated pp65 were processed more efficiently is unclear but this could be due to relatively unstable pp65 protein prone to degradation and entry into the processing pathway. Additional studies are now in progress to address the question.

One of the interesting findings of this study is that linear expression fragments encoding the right epitopes with only a single amino acid extension at the C-terminus were not recognized by each CTL line efficiently, whereas N-terminal extension rarely affected the recognition (Figure 4). Aminopeptidases, such as endoplasmic reticulum aminopeptidase 1 (ERAP1)^36-38  or leucine aminopeptidase,³⁹  may be able to trim N-terminal extensions and create peptides with optimal length for binding to major histocompatibility complex (MHC). However, mammalian cells lack carboxypeptidases,^39-41  thus proteasome is solely responsible for creating the correct C-terminus of the epitopes. Although a very limited number of amino acids adjacent to both sides of a proteasomal cleavage site contribute to cleavage site selection,⁴²  only a single amino acid extension at the C-terminus may be too short to be removed efficiently by proteasomes. In addition, constitutive proteasomes might be dominant in 293T cells⁴³  rather than immunoproteasomes, which have greater efficacy and are expressed in LCLs or mature dendritic cells. This may partially explain why C-terminal–extended epitopes were poorly recognized in our experiments.

We identified 2 new epitopes presented by HLA-A molecules. One is RIFAELEGV, dominantly presented by HLA-A*0207 but not by HLA-A*0201. This result underscores differential peptide repertories that bind to HLA-A*0207 and -A*0201, probably influenced by a single amino acid substitution at the floor of the binding groove.⁴⁴  The other is presented by HLA-A*1101 (ATVQGQNLK; aa's 501-509): this seems to be the dominant epitope presented by the allele, although a subdominant one may be located between aa 211 and aa 234. Both the dominant and undefined subdominant epitopes are, however, different from those reported previously, GPISGHVLK; aa 16-24.^9,30  For HLA-B alleles we identified 7 new epitopes, 4 of which are restricted by the HLA-B40 group. Among them, HERNGFTVI (aa's 267-275) is presented by both HLA-B*4001 and -B*4002. HLA-B*4001 presents an additional epitope, CEDVPSGKL (aa's 232-240). An HLA-B*4403–restricted epitope found in this study (SEHPTFTSQY; aa364-373) differs from the HLA-B44–restricted one reported earlier (EFFWDANDIY; aa's 512-521).^14,34  HLA-B*4402 and -B*4403 are 2 major HLA-B44 subtypes in white individuals,^45,46  and HLA-B*4403 is most frequent HLA-B44 subtype in Japanese individuals.^47-49  Because the reported epitope, EFFWDANDIY, was listed as the HLA-B*4402–restricted one,³⁵  these 2 epitopes might be restricted by a different subtype of HLA-B44. All the data imply that subtle differences in amino acid residues facing the groove of HLAs have an impact on the peptide binding and subsequent CMV-pp65–specific T-cell responses.

This paper describes, for the first time to our knowledge, pp65-specific epitopes presented by HLA-C alleles. A unique epitope, VVCAHELVC (aa's 198-206), was presented by both HLA-Cw*0801 and -Cw*1502. Since there seems to be no information on the peptide binding motif for these HLA alleles, the epitope was determined by gene engineering of pp65, followed by probing with CTL restricted to each HLA-C allele. The results should shed light on the structural basis of understanding the HLA-C molecules. Interestingly, an HLA-Cw*0401–restricted epitope, QYDPVAALF (aa's 341-349),¹⁵  is also dominantly presented by HLA-A*2402 allele. The binding motifs of those 2 alleles are similar to each other (ie, Tyr, Pro, or Phe at the second position and Leu or Phe at the C-terminus in HLA-Cw*0401; and Tyr at the second position, and Ile, Leu, or Phe at the C-terminus in HLA-A*2402). From the point of immunotherapy, this single peptide has great advantages among populations where HLA-A*2402 and -Cw*0401 are common.

There are only a few reports of successful generation of pp65-specific CTL from CMV-seronegative donors. Kleihauer et al⁵⁰  showed that cytotoxic T-cell lines were generated from 2 of 11 seronegative donors starting with 3 × 10⁶ PBMCs on stimulation with pp65 peptide–pulsed monocyte-derived dendritic cells. In our previous study, we failed to generate pp65-specific CTL lines from seronegative donors using pp65-transduced CD40-B cells as APCs, but in this study CTL lines that could lyse LCLs expressing endogenously processed peptides from the transduced pp65 gene were generated in 2 of 4 cases (N02 and N03) by using CD40-B cells pulsed with a mixture of the peptides as APCs. However CMV-infected fibroblasts were recognized weakly only by the CTL line from N02. Again, the better antigen processing by LCLs than fibroblasts might contribute to the better recognition by the CTL lines. Such CTLs may be able to only lyse CMV-infected hematopoietic cells in vivo. It is noted in this regard that CMV is reported to infect not only stroma and epithelial cells but also hematopoietic cells including CD34⁺ stem cells, monocytes, and dendritic cells.^51-55  In the other 2 cases, the CTL lines could recognize only peptide-pulsed LCLs. This observation seems to reflect the results of peptide titration experiments by ELISPOT assay; the peptide concentration to yield half-maximal spots was found to be 10- to 20-fold lower in N02 and N03 lines compared with N06 and N08 lines, suggesting the lower avidity of the lines generated from N06 and N08. To overcome this problem, initial stimulation with peptides at a lower concentration may induce higher avidity CTLs.⁵⁶  Alternatively, more T-cell input at the time of initial stimulation may be needed to induce CTL lines from rare precursor T cells with higher affinity T-cell receptor since estimated precursor frequency of naive T cells against a single epitope is very low (for example, one in 5 × 10⁶ of CD8⁺ cells for lymphocytic choriomeningitis virus in mice⁵⁷ ). Further efforts are now underway to establish better CTL induction conditions from seronegative donors.

In summary, we here identified 14 novel CTL epitopes derived from CMV-pp65 antigen restricted by HLA-A, -B, and -C alleles. These should be useful for immunologic monitoring of individuals expressing these HLA class I alleles and also for generation of pp65-specific CTLs from not only seropositive but also seronegative donors. In addition, our present approach of epitope identification applying deletion mutants and linear expression fragments, together with the efficient generation of CTL lines, may be applicable for other tumor-specific or viral antigens.

NIH-3T3-hCD40 ligand cells were kindly provided by Dr Gordon Freeman. Valuable discussions and suggestions by Drs T. Kiyono, A. Uenaka, Y. Nagata, M. Yazaki, T. Tsurumi, and E. Nakayama are highly appreciated. We are very grateful to Y. Matsudaira, K. Nishida, Y. Nakao, and H. Tamaki for their expert technical assistance.