Using synthetic peptides, the HLA-B27-restricted CTL response to EBV in asymptomatic virus carriers has been mapped to four epitope regions in EBV latent cycle Ags. One of these peptide-defined epitopes (RRIYDLIEL) tends to be immunodominant and is recognized in the context of all three B27 subtypes studied, B*2702, B*2704, and B*2705. The other peptide-defined epitopes induce responses only in the context of one subtype, the immunogenic combinations being RRARSLSAERY/B*2702, RRRWRRLTV/B*2704, and FRKAQIQGL/B*2705. We used immunoaffinity chromatography to isolate the naturally presented viral peptides associated with these MHC class I molecules on the surface of EBV-transformed B-LCL. Using CTL reconstitution assays in conjunction with mass spectrometry, we established that the naturally processed and presented peptides are identical with the previously identified synthetic sequences. Despite the subtype-specific immunogenicity of three of the four epitopes, all four epitope peptides were found in association with each of the three different HLA-B27 subtypes. Indeed, those peptides that failed to induce a response in the context of a particular HLA-B27 subtype were frequently presented at greater abundance by that subtype than were the immunogenic peptides. Furthermore, among the peptides that did induce a response, immunodominance did not correlate with epitope abundance; in fact the immunodominant RRIYDLIEL epitope was least abundant, being present at less than one copy per cell. The relationship of this unexpected finding to the persistence of EBV is discussed.
Epstein-Barr virus is a human γ Herpesvirus that infects the vast majority of the world’s population (reviewed in Ref. 1). This orally transmitted virus replicates in the oropharynx of infected individuals (2) but generalizes as a latent infection of the B lymphocyte pool through a process of virus-induced B cell growth transformation. These growth-transformed B cells, of which B-lymphoblastoid cell lines (B-LCL)3 are the in vitro counterpart, express eight latent cycle viral proteins: the nuclear Ags Epstein-Barr nuclear Ag (EBNA) 1, 2, 3A, 3B, 3C, and leader protein (LP) and the membrane proteins latent membrane protein (LMP) 1 and 2 (3). CTL surveillance directed against such latently infected B cells is likely to be important in the control of EBV infection (4). Certainly EBV-specific CTL can readily be reactivated in vitro from healthy viral carriers by stimulating their peripheral blood T cells with the autologous B-LCL (5, 6). Therefore, determining those CD8+ CTL epitopes responsible for this T cell-mediated immunity is of critical importance in the design of appropriate immunotherapy for the treatment of EBV-associated malignancies.
CD8+ CTL recognize peptides associated with class I MHC molecules on the surface of transformed and virally infected cells. These peptides are derived from proteins degraded in the cytosol by the proteasome and shuttled into the endoplasmic reticulum (ER) via TAP. Once these peptides enter the ER, they associate with MHC class I H chains and β2-microglobulin to form a trimeric complex that proceeds through the secretory pathway to the cell surface (reviewed in Ref. 7). Numerous reports have described the use of synthetic peptides to identify the epitopes recognized by CD8+ CTL specific for latent stage EBV-infected cells (reviewed in Ref. 8). Studies involving a range of different HLA class I alleles have demonstrated strong responses against epitopes derived from the EBNA3A, EBNA3B, and EBNA3C proteins, quite often accompanied by subdominant responses to LMP2 epitopes. By contrast, EBNA2-, EBNA-LP-, and LMP1-specific responses are less frequent while CTL reactivity toward EBNA1 is apparently rare. Precise epitope choice depends upon the identity of the MHC class I-restricting allele, and one of the most closely studied sets of alleles is the HLA-B27 family (9, 10). CTL specific for the peptide RRIYDLIEL (from EBNA3C) were found in individuals of three different subtypes of HLA-B27 (B*2702, B*2704, and B*2705; Ref. 11) and tend to constitute the strongest response in such individuals. In addition, CTL specific for the peptides RRARSLSAERY (from EBNA3B), RRRWRRLTV (from LMP2), and FRKAQIQGL (from EBNA3C) were found only in individuals of the HLA-B*2702, B*2704, and B*2705 subtypes, respectively (9, 10, 11). Interestingly, however, these four synthetic peptides all contain an arginine at position 2 and a tyrosine or hydrophobic residue at the C terminus, thereby conforming to the previously described binding motif for HLA-B27 molecules (12, 13, 14). Furthermore, when assayed by peptide-induced stabilization of HLA-B27 molecules at the cell surface, the pan-B27 epitope RRIYDLIEL and also the two subtype-specific epitopes tested, RRARSLSAERY and RRRWRRLTV, appeared to bind with similar affinities to HLA-B*2702, B*2704, and B*2705 (15). The present work was prompted by these apparent discrepancies between the capacity of individual EBV peptides to bind different HLA-B27 subtypes in vitro and their capacity to induce CTL responses in the context of these subtypes in vivo.
There are a number of possible explanations for the lack of immunogenicity of the RRARSLSAERY, RRRWRRLTV, and FRKAQIQGL peptides in the context of certain HLA-B27 subtypes. Although the synthetic peptide approach has been widely used to identify CTL epitopes, there have been instances where the sequences of the predicted synthetic peptides did not correspond to the sequences of the naturally processed and presented epitopes due to alterations in the length of or posttranslational modifications of the natural epitopes (16, 17, 18, 19). A second possibility was that despite the presence of an appropriate binding motif and the ability to bind all three HLA-B27 subtypes efficiently in vitro, these peptides were not presented by certain HLA-B27 subtypes in vivo. A third possibility was that the differences in immunogenicity reflected the level of presentation of each of these epitopes by different HLA-B27 molecules. To address these possibilities, we isolated the peptides associated with the HLA-B*2702, B*2704, and B*2705 molecules from the surface of EBV-transformed B-LCL by immunoaffinity chromatography. The sequence identity and abundance of each naturally presented EBV epitope were established using CTL reconstitution assays in combination with previously described mass spectrometric techniques (20, 21).
Materials and Methods
EBV-transformed B-LCL were generated from the following EBV+ donors by infecting PBMC with the B95.8 EBV strain in vitro: DH (HLA-A2, A11, B27.04, B40), KOR (HLA-A24, A28, B27.02, B44), LY (HLA-A1, A24, B27.02, B35), RT (HLA-A2, A24, B27.05, B35), and SC (HLA-A2, A2, B27.05, B27.05). The B95.8 strain encodes sequences identical with those contained in the four HLA-B27-restricted synthetic peptides, and each of these B95.8-transformed B-LCL was recognized by appropriate HLA-B27-restricted CTL (11). A spontaneous B-LCL was generated from donor NW (A3, A31, B8, B27.02) by transformation with the donor’s endogenous viral isolate (identified as NW spont LCL). B-LCL were cultured in RPMI 1640 supplemented with 5% FBS containing SerXTend (Irvine Scientific, Santa Ana, CA), 2 mM glutamine, and 15 mM HEPES. C1R-B2702, C1R-B2704, and C1R-B2705 are transfectants of the HLA-A- and B-negative cell line C1R. C1R-B2702 and C1R-B2704 were obtained from Dr. Robert Colbert (University of Cincinnati) and were maintained in the above culture medium supplemented with 500 μg/ml G418. C1R-B2705 was obtained from Dr. Peter Cresswell (Yale University, New Haven, CT) and was maintained in the above culture medium supplemented with 300 μg/ml G418.
EBV-specific CTL clones were derived and grown as described previously (11). Briefly, PBMC from the above donors were cultured with gamma-irradiated B95.8 EBV-transformed autologous B-LCL at a responder-to-stimulator ratio of 40:1. CTL clones were derived from these activated populations by seeding in semisolid agarose or by limiting dilution cloning. CTL clones were maintained by weekly stimulation with gamma-irradiated autologous B-LCL in RPMI 1640 supplemented with 10% FBS, 2 mM glutamine, 15 mM HEPES, 1% human serum, 25% supernatant of the IL-2-producing MLA-144 cell line, and 50 Us/ml rIL-2 (Chiron, Emeryville, CA).
Peptides were synthesized by using solid-phase F-moc methodology using Gilson Medical Electronics (Middleton, WI) AMS422 peptide synthesizers. Peptides were purified to >90% homogeneity by reversed-phase HPLC, and their identities were confirmed by mass spectrometric analysis. The deuterium-labeled amino acid leucine, d10 (Cambridge Isotope Laboratories, Andover, MA), was used to synthesize the RRIYD*LIEL peptide standard for quantitation.
Extraction and separation of HLA-B27-associated peptides
HLA-B*2702, B*2704, or B*2705 molecules were immunoaffinity purified from 1010 B-LCL using the mAb ME1–1.2 (a generous gift of Dr. Charles Lutz, University of Iowa, Des Moines, IA), and the bound peptides were eluted with acid (20). In some experiments, peptide extracts were fractionated by HPLC on a Brownlee reversed-phase C18 column (7 μm particles, 30 nm pore size, 2.1 mm i.d., 3 cm length) (Varian, Walnut Creek, CA) using a gradient of 0% buffer B (0–5 min), and 0% buffer B to 100% buffer B in 40 min, at a flow rate of 200 μl/min. Buffer A was 0.1% trifluoroacetic acid (TFA) in water, and buffer B was 0.085% TFA in 60% acetonitrile, 40% water. Beginning at 5 min, 200 μl fractions were collected into Eppendorf tubes (Sarstedt, Newton, NC). Other peptide extracts were fractionated by HPLC on a Higgins Analytical reversed-phase C18 column (5 μm particles, 30 nm pore size, 2.1 mm diameter, 4 cm long) (Bodman, Ashton, PA) using the same buffers and gradient as described above. Beginning at 5 min, 140 μl fractions were collected into Eppendorf tubes (Sarstedt).
CTL epitope reconstitution assays
A total of 2000 51Cr-labeled targets (indicated in each figure) were incubated with either 5–10% of each HPLC fraction, corresponding to 5 × 108 to 1 × 109 cell equivalents, or graded doses of synthetic peptides in 150 μl of HBSS supplemented with 1% BSA, 2 mM glutamine, and 30 mM HEPES for 2–3 h at room temperature. CTL were added at E:T of 10:1 or 20:1, and 51Cr release was determined after 4–5 h at 37°C.
Assays for epitope immunodominance
CTL clones were screened in standard 4- to 5- h 51Cr release assays against autologous target cells either overexpressing individual EBV latent Ags from recombinant vaccinia virus vectors or preincubated with synthetic peptide epitopes as previously described (11). T cells within fresh PBMC populations specific for defined EBV-derived epitopes were detected using an enzyme-linked immunospot assay for single-cell IFN-γ release as described (22).
Tandem mass spectrometry (MS/MS)
First-dimension HPLC fractions of extracted peptides corresponding to 5 × 107 to 5 × 108 cell equivalents were analyzed on an LCQ ion trap mass spectrometer (Finnigan, San Jose, CA) equipped with a sheathless nano-HPLC microelectrospray ionization source as previously described (21). Nano-HPLC columns were constructed of 75 μm i.d.-fused silica (Polymi cro Technologies, Phoenix, AZ) packed with 10–12 cm of 5 μm diameter C18 packing material (Waters, Milford, MA). For HPLC, the mobile phase consisted of solvent A, 0.1% acetic acid in water, and solvent B, 0.1% acetic acid in 60:40 acetonitrile:water. Peptides were eluted with a gradient consisting of 0–35% B in 33 min, 35–80% B in 4 min, 80–0% B in 3 min, followed by a 7 min wash with 100% A. Mass spectrometric data were acquired by manually switching from MS-only mode to collision-activated dissociation (CAD) mode after elution of a marker peptide. MS/MS spectra were acquired using three microscans with a maximum injection time of 500 ms/microscan, a 1.6 amu isolation window, and 30% relative collision energy. For each epitope, the 1.6 amu window was centered on the m/z ratio corresponding to the observed dominant charge state of the peptide. Synthetic peptides were analyzed under the same conditions.
Quantitation of the naturally processed RRIYDLIEL epitope was performed using a home-built Fourier transform ion cyclotron resonance mass spectrometer (21), interfaced with the same nano-HPLC microelectrospray ionization source described above. For HPLC, the mobile phase consisted of solvent A, 0.1% acetic acid in water, and solvent B, 0.1% acetic acid in 60:40 acetonitrile:water. Peptides were eluted with a gradient consisting of 0–40% B in 20 min, 40–100% B in 4 min, 100–0% B in 3 min, followed by a 7-min wash with 100% A. High-resolution (m/Δm ∼35,000) mass spectra were recorded at a rate of ∼1/s. A SWIFT (23) waveform with a notch of zero excitation energy centered about the mass range, 591–606, was used to isolate naturally processed RRIYDLIEL and the deuterated analogue. Ejection of all other species from the Fourier transform mass spectrometer Penning trap before detection minimized deleterious mass shifts resulting from excess space charge. Broadband excitation of the remaining ions was followed by heterodyne detection over the mass range 470–700. Naturally processed RRIYDLIEL was quantitated by comparing its signal intensity with that for a deuterium-labeled standard. Analysis of various concentrations of the deuterated synthetic peptide RRIYD*LIEL (*L = leucine, d10) produced a linear standard curve (R2 = 0.9996) spanning the range from 10 amol to 5 fmol.
Hierarchy of CTL reactivities in HLA-B27+, EBV+ donors
The epitope specificity of HLA-B27-restricted CTL from two donors each of the HLA-B*2702 and HLA-B*2705 subtypes and from one donor of the HLA-B*2704 subtype was analyzed after stimulation with B95.8 EBV-transformed autologous B-LCL (Table I). From all donors of all three HLA-B27 subtypes, CTL clones were isolated that recognized the peptide RRIYDLIEL restricted by the appropriate subtype. This was the dominant reactivity in the majority of donors tested. In contrast, recognition of the remaining three EBV peptides depended on the subtype of the individuals from which CTL clones were isolated. In the HLA-B*2702+ donors, clones reactive with the RRARSLSAERY peptide were isolated, but there were no detectable CTL responses to RRRWRRLTV or FRKAQIQGL. In the HLA-B*2705+ donors, FRKAQIQGL was recognized while RRARSLSAERY and RRRWRRLTV were not. Finally, from the HLA-B*2704+ donor DH, CTL clones reactive with the RRRWRRLTV peptide were isolated, while CTL responses to RRARSLSAERY or FRKAQIQGL were not detected. When the PBMC from these donors were tested immediately ex vivo for epitope-specific reactivities in rapid enzyme-linked immunospot assays of peptide-induced IFN-γ release, we observed the same pattern of results (data not shown). This confirmed the relative immunodominance of the pan-B27-restricted RRIYDLIEL epitope and the subtype-restricted nature of the other epitopes, as originally determined by in vitro outgrowth analysis. These results suggested that the lack of immunogenicity of two of the four EBV peptides in individuals of a particular HLA-B27 subtype might be due to either the differences in structure between the synthetic peptides and the naturally processed epitopes, leading to a failure of recognition, or to the absence of these epitopes on the surface of B-LCL in the context of “nonimmunogenic” HLA-B27 subtypes. To test these hypotheses, we identified the naturally processed forms of these epitopes and analyzed their expression in association with each of the HLA-B27 subtypes on B95.8 EBV-transformed B-LCL.
|Donor .||Subtype .||No. of Expts. .||No. of Clones Specific for .||.||.||.|
|.||.||.||RRIYDLIEL .||RRARSLSAERY .||RRRWRRLTV .||FRKAQIQGL .|
|Donor .||Subtype .||No. of Expts. .||No. of Clones Specific for .||.||.||.|
|.||.||.||RRIYDLIEL .||RRARSLSAERY .||RRRWRRLTV .||FRKAQIQGL .|
CTL clones were generated either by limited dilution cloning or by seeding in semisolid agarose. Proliferative CTL clones were tested in 51Cr release assays for their ability to recognize autologous targets pulsed separately with each of the four synthetic peptides.
The clonal analysis data from donors LY and KOR has been previously reported (9).
NT, Not tested. Donor KOR was only screened against a limited panel of peptides.
Identification of the naturally processed epitope recognized by an RRARSLSAERY-reactive, HLA-B*2702-restricted CTL clone
We first determined the identity of the naturally processed and presented epitope recognized by HLA-B*2702-restricted CTL reactive with the EBNA3B-derived peptide RRARSLSAERY. HLA-B*2702 molecules on LY LCL were immunoaffinity purified, and the endogenous peptides associated with this class I MHC molecule were extracted. Ninety percent of this extract was separated by reversed-phase HPLC using TFA as the ion-pairing agent. Aliquots of individual HPLC fractions were incubated with C1R-B2702 cells to test for the presence of a peptide capable of binding to HLA-B*2702 and reconstituting the epitope for the HLA-B*2702-restricted, RRARSLSAERY-specific CTL clone LYcl48. Reconstituting activity was found in fractions 26 and 27 (Fig. 1,A). To determine the identity of the peptide responsible for biological activity in the HLA-B*2702+ extract, the remaining 10% of the extract that had not been fractionated was spiked with 12.5 pmol synthetic RRARSLSAERY and chromatographed under identical conditions to the 90% described above. Reconstitution assays were conducted on this second set of fractions. We found two fractions of reconstituting activity that coeluted with those active fractions from the naturally processed material (Fig. 1 A). It should be noted that the reconstituting activity in the spiked samples was achieved using only one-tenth the amount of the naturally processed material and therefore is largely or entirely attributable to the synthetic peptide. These results were consistent with the idea that the naturally processed peptide was identical in structure to the synthetic material.
The presence of the RRARSLSAERY peptide in the fractions with reconstituting activity was confirmed by mass spectrometric comparison of the synthetic peptide with ions in these fractions. In the electrospray ionization mass spectrum of synthetic RRARSLSAERY, the most abundant (base peak) ion is observed at m/z 342.1 (average mass) and corresponds to (M + 4H)+4 (data not shown). In the CAD mass spectrum recorded on the m/z 342.1 ions (Fig. 2,A), the base peak ion (m/z 301) corresponds to the loss of the C-terminal Tyr as a neutral residue from the (M + 4H)+4 parent ion. Residues 1 and 5–10 were defined by a series of singly, doubly, and triply charged ions of type b, and residues 3–9 were defined by a series of singly and doubly charged ions of type y. Ions of m/z 342.1 (average mass) were also observed in the HPLC fractions that reconstitute the epitope for HLA-B*2702-restricted, RRARSLSAERY-specific CTL clone LYc148. The CAD spectrum of these ions (Fig. 2,B) was recorded on sample at the low attomole level and thus is contaminated with background ions and missing several of the low abundant fragments observed in Fig. 2,A. Nonetheless, the dominant ion in Fig. 2,B (m/z 301) is identical with that in Fig. 2 A and corresponds to loss of C-terminal Tyr as a neutral residue from the (M + 4H)+4 parent ion. Also observed are doubly charged ions of type b and singly charged ions of type y that define the last five residues in the peptide as SAERY. Fourier transform MS was employed to assign the monoisotopic mass of the (M + 4H)+4 parent ion in the above HPLC fractions as 341.939. This is in excellent agreement with the theoretical mass, 341.943, calculated for the (M + 4H)+4 ion of RRARSLSAERY. Taken together, the above data provide strong support for the presence of naturally processed RRARSLSAERY in the HPLC fractions recognized by CTL clone LYc148.
One hypothesis to explain the failure to generate RRARSLSAERY-reactive CTL from HLA-B*2704+ or B*2705+ individuals was that these two subtypes did not present a naturally processed peptide corresponding to this epitope in vivo. Accordingly, we prepared peptide extracts from the immunoaffinity purified HLA-B*2704 and HLA-B*2705 molecules on DH LCL and RT LCL, respectively, and fractionated them by HPLC. When unspiked HPLC fractions of these extracts were incubated with NW spont LCL, reconstituting activity for the LYcl48 CTL clone was found in single fractions of each (Fig. 1, B and C). Again, the activities identified in these unspiked fractions coeluted with activities found in RRARSLSAERY “spiked” peptide extracts. Finally, mass spectrometric analysis of the peptides in the active fractions of the unspiked extracts identified a mass of 342.1 that gave a CAD spectrum identical with synthetic RRARSLSAERY (data not shown). These results indicate that the EBNA3B-derived peptide RRARSLSAERY is processed and presented on the surface of B95.8 EBV-transformed B-LCL of the HLA-B*2702, B*2704, and B*2705 subtypes. Therefore, the failure to generate RRARSLSAERY-reactive CTL from HLA-B*2704+ or B*2705+ individuals is not due to the absence of this epitope in vivo.
Identification of peptides recognized by HLA-B*2705-restricted, FRKAQIQGL-reactive and HLA-B*2704-restricted, RRRWRRLTV-reactive CTL clones
Similar to the situation with the RRARSLSAERY epitope, the results summarized in Table I had established that reactivity against the EBNA3C-derived peptide FRKAQIQGL is present only in EBV-specific CTL clones from HLA-B*2705+ individuals. HPLC fractions containing reconstituting activity for the HLA-B*2705-restricted, FRKAQIQGL-reactive CTL clone SCcl30 were identified in the unspiked extract of HLA-B*2705, as well as extracts from B*2702 and B*2704 (Fig. 3). The active fractions all coeluted with synthetic FRKAQIQGL (Fig. 3), and the presence of this peptide was confirmed by mass spectrometry (data not shown). Thus, FRKAQIQGL is processed and presented on the surface of B95.8 EBV-transformed B-LCL of all three HLA-B27 subtypes, despite the fact that specific CTL for this epitope are demonstrable only in HLA-B*2705+ individuals.
Finally, although CTL reactive with the LMP2-derived peptide RRRWRRLTV are found only in HLA-B*2704+ individuals, reconstituting activity for the HLA-B*2704-restricted, RRRWRRLTV-reactive CTL clones DHcl15 and DHcl1 was found in fractions from both HLA-B*2702+ and HLA-B*2705+ peptide extracts (Fig. 4). The active fractions coeluted with synthetic RRRWRRLTV (Fig. 4), and the presence of this peptide was confirmed by mass spectrometry (data not shown). Surprisingly, however, activity was not evident in unspiked peptides extracted from HLA-B*2704 itself. We failed to identify active fractions from multiple peptide extracts from HLA-B*2704+ DH LCL using either of the HLA-B*2704-restricted, RRRWRRLTV-specific CTL clones (Fig. 4,B and data not shown). However, the peptide RRRWRRLTV was detected by mass spectrometry in fraction 17 in the peptide extract from HLA-B*2704. Shown in Fig. 5,A is the CAD spectrum recorded on ions of m/z 325.7 (average mass), corresponding to (M + 4H)+4 from the synthetic peptide, RRRWRRLTV. Abundant ions in the spectrum at m/z 428.1, 300.9, and 275.6 result from the loss of NH4+, C-terminal Val as a neutral residue, and both Val and Thr as neutral residues, respectively. Losses of both NH3 and guanidine also occur from the side chains of Arg residues present in the above ions. In the CAD spectrum recorded on ions of m/z 325.7 (average mass) in fraction 17 of the peptide extract from HLA-B*2704, the chemical background is high because the data was recorded on sample at the low attomole level (Fig. 5,B). Nevertheless, the same fragment ions observed in Fig. 5,A also appear in Fig. 5 B. Fourier transform mass spectrometry was employed to assign the monoisotopic mass of the (M + 4H)+4 parent ion in the above HPLC fraction as 325.455, in excellent agreement with the theoretical mass, 325.457, calculated for the (M + 4H)+4 ion of RRRWRRLTV. Taken together, the above results provide strong support for the conclusion that RRRWRRLTV is indeed present in HPLC fraction 17. These results indicated that the RRRWRRLTV peptide is processed and presented on the surface of B95.8 EBV-transformed B-LCL of the HLA-B*2702, B*2704, and B*2705 subtypes, despite the fact that CTL against it are only demonstrable in HLA-B*2704+ individuals. However, the difficulty in demonstrating this epitope by CTL reconstitution in extracts of HLA-B*2704+ cells suggested either that it coeluted with a competitor peptide or was present in lower amounts in this extract.
Quantitation of nonimmunogenic HLA-B27 associated EBV epitopes
The foregoing results established that the EBV-derived peptides RRARSLSAERY, FRKAQIQGL, and RRRWRRLTV were all present on the surface of HLA-B*2702+, B*2704+, and B*2705+ B-LCL, despite the fact that each is immunogenic only in the context of one of these subtypes. We next hypothesized that the immunogenic epitopes would be expressed on the cell surface at relatively high copy number, while the nonimmunogenic epitopes would be present in very small amounts. To test this hypothesis, we quantitated the amount of peptide present in the extracts based on the amounts required to give lysis comparable to that of a known amount of synthetic peptide. Based on the surface expression kinetics4 and the intracellular transit time (24) of HLA-B27 in B-LCL, we estimate that 90% of the peptides in these extracts are derived from HLA-B27 molecules that are on the cell surface at the time of extraction. In the example shown in Fig. 6, incubation of C1R-B2705 targets with 1 μl of HPLC fraction 17 from DH LCL resulted in 40% specific lysis, which was comparable to that achieved by incubation with 3 × 10−3 nM of the synthetic peptide FRKAQIQGL. Based on a total HPLC fraction volume of 200 μl from the peptide extract of 1010 DH LCL, this corresponds to 1.5 × 10−4 nmol/1010 cells, or ∼9 copies/cell. Previously, we had determined that the percent recovery of the naturally occurring FRKAQIQGL peptide through the final steps of our peptide extraction and HPLC separation procedures was ∼48% (data not shown); thus the corrected abundance of the FRKAQIQGL epitope on the surface of HLA-B*2704+ B-LCL is ∼19 copies/cell.
Surprisingly, the results demonstrated that the nonimmunogenic epitopes were of equal or greater abundance than the immunogenic epitopes (Table II). For example, in the peptides extracted from HLA-B*2702+ LY LCL, the nonimmunogenic FRKAQIQGL and RRRWRRLTV peptides were 6- to 36-fold more abundant than the immunogenic RRARSLSAERY epitope. Likewise, from HLA-B*2704+ DH LCL, the nonimmunogenic FRKAQIQGL and RRARSLSAERY peptides were 20- to 30-fold more abundant that the immunogenic RRRWRRLTV epitope. Finally, all three EBV-derived epitopes were equally abundant on the surface of the HLA-B*2705+ RT LCL. Therefore, the failure to generate CTL specific for all four EBV epitopes in individuals of all three HLA-B27 subtypes was not due to low copy numbers of these peptides on the surface of B95.8 EBV-transformed B-LCL because their density was equal to or greater than the abundance of the immunogenic epitopes.
|Donor .||Subtype .||Peptide Copies/Cell .||.||.|
|.||.||RRARSLSAERY .||RRRWRRLTV .||FRKAQIQGL .|
|Donor .||Subtype .||Peptide Copies/Cell .||.||.|
|.||.||RRARSLSAERY .||RRRWRRLTV .||FRKAQIQGL .|
Values shown are the mean of one to three experiments. For each peptide, a percent recovery has been determined and used in the calculations of epitope density to account for peptide losses that occur during the peptide extraction and HPLC separation procedures.
This epitope could not be identified in the DH peptide extract by CTL analysis, although its presence was established by MS. HLA-B*2704-restricted, RRRWRRLTV-specific CTL were able to detect synthetic peptide at the lowest concentration of 1 pg/mL. The cell surface density of this peptide was determined by MS to be one copy per cell.
Identification and quantitation of the naturally processed epitope recognized by HLA-B27-restricted, RRIYDLIEL-reactive CTL
Given the results above, we turned our attention to the EBNA3C-derived epitope represented by the RRIYDLIEL peptide, which is recognized by CTL from individuals of all three HLA-B27 subtypes. The endogenously processed peptides corresponding to this epitope were identified in HPLC fractions of peptides extracted from HLA-B*2702, HLA-B*2704, and HLA-B*2705 as above. Fractions were incubated either with C1R-B2702 targets and the HLA-B*2702-restricted, RRIYDLIEL-reactive CTL clone LYcl74 or C1R-B2705 targets and the HLA-B*2705-restricted, RRIYDLIEL-reactive CTL clone RTcl5. In each case, epitope reconstituting activity was found in a single peak of one to two HPLC fractions and coeluted with synthetic RRIYDLIEL (Fig. 7). However, in the HLA-B*2705+ peptide extract, a second peak of activity was detected in fraction 26; this additional peak was seen in multiple experiments, but the epitope responsible for this activity has not yet been identified. Mass spectrometry of the peptides in the unspiked active fractions of the HLA-B*2702+ and HLA-B*2704+ extracts as well as unspiked HPLC fractions 22 and 23 from the HLA-B*2705+ extract confirmed the presence of naturally processed RRIYDLIEL (data not shown). These results indicate that this peptide is presented on the surface of B95.8 EBV-transformed B-LCL expressing all three HLA-B27 subtypes.
The amounts of this peptide expressed in association with the different HLA-B27 subtypes were determined based on the amounts of each of the active unspiked HPLC fractions required to give lysis comparable to that of a known amount of synthetic RRIYDLIEL. We found that endogenously processed RRIYDLIEL was presented at less than one copy per cell in the context of all three HLA-B27 subtypes (Table III). This is an extremely low abundance for any epitope, and it is also significantly less abundant than the nonimmunogenic epitopes presented by each subtype. Because of the extremely low abundance of RRIYDLIEL established using CTL, we were concerned that this number might reflect disproportionate loss of the peptide during the extraction and fractionation procedure. Accordingly, we quantitated the peptide by a rigorous approach using deuterium-labeled RRIYD*LIEL as an internal standard in conjunction with mass spectrometry. Deuterium-labeled and naturally occurring RRIYDLIEL will behave identically during the extraction and will coelute during HPLC fractionation. However, deuterium-labeled RRIYD*LIEL differs in mass from the naturally processed form by 10 amu and thus is easily distinguishable by mass spectrometry. By doping a known quantity of deuterium-labeled RRIYD*LIEL synthetic peptide onto the immunoaffinity column immediately before acid elution of the peptide-MHC complexes from the column, we determined that the percent recovery of the naturally occurring RRIYDLIEL peptide through the final steps of our peptide extraction and HPLC separation procedures was ∼55%. The results of this mass spectrometric analysis confirmed the abundance of the natural RRIYDLIEL epitopes previously determined by CTL analysis (Table III). Thus, based on two independent methods of epitope quantitation, we conclude that the immunodominant RRIYDLIEL epitope is presented by all three HLA-B27 subtypes at less than one copy per cell.
|Donor .||Subtype .||Epitope Density (copies/cell) as Determined by .||.|
|.||.||CTL analysisb .||MSc .|
|Donor .||Subtype .||Epitope Density (copies/cell) as Determined by .||.|
|.||.||CTL analysisb .||MSc .|
Values were calculated as described in Results and account for losses during the peptide extraction and HPLC separation procedures.
Values are the mean of two experiments from one peptide extract of each LCL.
Values are the mean of multiple analyses from the same peptide extract of each LCL.
In this study, we characterized the expression of four peptide epitopes from the EBV latent proteins EBNA3B, EBNA3C, and LMP2 that are naturally presented on the surface of HLA-B*2702+, B*2704+, or B*2705+ B95.8 EBV-transformed B-LCL. Although peptides corresponding to all four epitopes were presented on the surface of B-LCL of the three different HLA-B27 subtypes, CTL reactivity toward each peptide differed among individuals of each subtype. For example, CTL that could recognize the RRARSLSAERY peptide were only found in HLA-B*2702+ individuals. Similarly, CTL reactivity to RRRWRRLTV was confined to HLA-B*2704+ individuals, and reactivity to FRKAQIQGL was only found in those who were HLA-B*2705+. In contrast, CTL specific for the RRIYDLIEL peptide could be generated from individuals with any of these subtypes. Upon quantitation of these epitopes using CTL reconstitution assays and mass spectrometry, we found that the density of the immunogenic epitopes for each HLA-B27 subtype was less than or equal to the abundance of the nonimmunogenic epitopes for that particular HLA-B27 subtype.
The use of HLA-B27-restricted CTL clones specific for the four different EBV synthetic peptides allowed us to identify HPLC fractions from peptide extracts of B95.8 EBV-transformed B-LCL of all three HLA-B27 subtypes that reconstituted activity for these CTL. The sequences of the natural peptides extracted from B-LCL and responsible for reconstituting activity for EBV-specific CTL were established by mass spectrometry to be identical with the sequences of the proposed synthetic peptides (9, 10, 11). The use of this technology, which has the capability of detecting modifications of peptide Ags, allowed us to definitely identify the correct structure of the naturally processed and presented epitopes. Furthermore, this is one of the first reports of the direct identification of viral peptides using mass spectrometry (19, 25).
By employing two independent means of epitope quantitation, we determined that the density of the RRIYDLIEL peptide on the surface of B95.8 EBV-transformed B-LCL of three different HLA-B27 subtypes was less than one copy per cell. This is the first direct demonstration of the expression of a CTL epitope at a cell surface density of less than one copy per cell. In that context, there is no reason to believe that the four B-LCL examined are in any way atypical. Expression levels of the EBNA3C protein (from which RRIYDLIEL is derived) were standard, as were the levels of HLA-B27 molecules present at the cell surface, and the total yields of HLA-B27-associated peptides obtained following immunoaffinity purification (data not shown). Previous studies have generally shown that the number of peptide-MHC complexes per target cell required for recognition and cytolysis by differentiated CTL varies from several thousand to <10 (26, 27, 28, 29, 30). Sykulev et al. previously provided more indirect evidence that in target cells pulsed with an average of three peptide epitope molecules per cell, cells expressing a single peptide-MHC complex could be recognized by CTL (31). Consistent with our quantitation is the observation that unmanipulated HLA-B27+ B-LCL are poorly recognized by RRIYDLIEL-specific CTL, even though these effectors are of extremely high avidity (SD50 = 1–5 pM) based on peptide dose-response curves (data not shown). Therefore, although the RRIYDLIEL peptide is present at less than one copy per cell on the surface of B-LCL, its density is still sufficient to allow for recognition, albeit poor, and lysis of B-LCL by RRIYDLIEL-specific CTL. Because it is generally believed that effective T cell activation requires cross-linking contingent on the engagement of more than one TCR, the question remains whether CTL are truly able to recognize target cells expressing only a single peptide epitope. One possibility is that other accessory molecules participate in signaling events in the immunological synapse and so substitute for the cross-linking of multiple TCR. Alternatively, at an average of one copy per cell, Poisson statistics predicts that 23% of the cells express two or more copies per cell at any point in time. It is possible that only these cells are in fact CTL targets. However, in a population, the exact cells expressing more than one epitope per cell would change with time, allowing a larger fraction of that population to ultimately be recognized.
Only a limited number of studies have examined the role of the abundance of an epitope and its immunogenicity by quantitating the naturally presented peptide Ags on the surface of an APC. Previously, several groups had shown that the magnitude of a T cell response to a peptide epitope was proportional to the density of that epitope (32, 33, 34, 35). In particular, Levitsky et al. demonstrated that an immunodominant HLA-A11-restricted CTL epitope from the EBV latent protein EBNA3B was more abundant than a subdominant epitope (36). Here we report an inverse relationship between the immunogenicity of multiple EBV-derived epitopes and their density on the surface of HLA-B27+ B-LCL. In another viral system, attempts to reactivate in vitro HLA-A*0201-restricted CTL specific for a measles virus epitope have not been successful although this peptide is highly abundant on the surface of measles-infected cells (37). Additionally, immunization of HLA-A2/Kb transgenic mice with high doses of this peptide failed to elicit a CTL response, whereas the administration of a 100-fold lower dose of the peptide proved immunogenic in vivo, suggesting that high doses of Ag inhibit the development of a CTL response to this peptide (37). Furthermore, in a study with the melanoma tumor Ags, immunization of HLA-A2 transgenic mice with large concentrations of tyrosinase elicited only a small population of CD8+ T cell responses, suggesting that the optimal dose of this tumor Ag for activating T cells is not necessarily the largest dose (38). Finally, Pamer and colleagues have also demonstrated an inverse relationship between the cell surface density of CTL epitopes from the intracellular bacterium Listeria monocytogenes and their immunogenicity (39, 40). Specifically, this group showed that a peptide from the listeriolysin O protein was less abundant than two peptides from the p60 protein. However, the listeriolysin O epitope elicited the largest primary and memory T cell responses during Listeria infection. These authors suggested that the immunogenicity of a particular epitope does not directly reflect the relative density of that Ag, but rather additional factors such as TCR repertoire and T cell affinity are likely to play a role in influencing the magnitude of the T cell response. Similar phenomena may explain the differences in immunogenicity that we observed among the various HLA-B27-restricted EBV epitopes.
An alternative model to explain our findings with these HLA-B27-restricted peptides is based on the fact that EBV establishes a persistent infection within its host. In other persistent viral infections such as lymphocytic choriomeningitis virus (41, 42, 43), hepatitis B virus (44, 45), and HIV (46, 47), it has been demonstrated that virus-specific, CD8+ T cells undergo repeated activation that ultimately results in activation-induced cell death (48). During a chronic EBV infection, it is possible that virus-specific CTL frequently encounter their cognate Ags on the surface of EBV-transformed B cells and that the continuous exposure of these CTL to EBV epitopes eventually results in activation-induced cell death and their disappearance from the memory T cell population. Because activation-induced cell death is dependent on encounter with Ag, presentation of a higher density of any individual virus-derived epitope on the surface of an EBV-transformed B-LCL could increase the probability of early activation, but ultimately deletion of, larger numbers of CTL specific for that epitope. Importantly, high-density epitope presentation is likely to be particularly effective in inducing the death of the highest avidity CTL (49). In this model, the apparent absence of a memory CTL response to “nonimmunogenic” epitopes would not in fact be due to an intrinsic lack of immunogenicity, but rather a consequence of the high density of these peptides presented on the surface of B-LCL in the context of certain HLA-B27 subtypes. Conversely, the CTL response to the immunodominant RRIYDLIEL peptide would persist in individuals of all three HLA-B27 subtypes because the low cell surface density of this epitope limits the opportunities for Ag encounter. Finally, this model predicts that immune responses against the higher density “nonimmunogenic” and “subtype specific” peptides would be observed in newly infected individuals expressing any of the three HLA-B27 subtypes.
In conclusion, here we identify four peptide epitopes from three EBV latent proteins that are naturally presented on the surface of HLA-B*2702+, B*2704+, or B*2705+ EBV-transformed B-LCL. In the context of HLA-B27, we also determined that higher cell-surface densities of these EBV-derived epitopes were correlated with an inability to demonstrate specific memory CTL in asymptomatic individuals. This relationship may be the means by which viruses such as EBV achieve a balance that enables them to persist despite an effective host immune response.
We thank Megan Partridge and Theresa Rodriguez for their expert technical assistance. We thank Dr. C. A. Mosse and Dr. C. J. Luckey for their helpful discussions.
This work was supported by U.S. Public Health Service Grants AI 20963 and AI 21393 (to V.L.C. and V.H.E.), National Institutes of Health Grants 2 R37-AI3393-06 (to R.E.C., J.S., R.E.S., J.A.M., F.M.W., and D.F.H.), and the Medical Research Council (to J.M.B. and A.B.R.).
Abbreviations used in this paper: B-LCL, lymphoblastoid cell line; EBNA, Epstein-Barr nuclear Ag; LP, leader protein; LMP, latent membrane protein; ER, endoplasmic reticulum; i.d., internal diameter; TFA, trifluoroacetic acid; MS/MS, tandem mass spectrometry; CAD, collision-activated dissociation.
C. J. Luckey, J. A. Marto, M. Partridge, E. Hall, F. M. White, J. D. Lippolis, J. Shabanowitz, D. F. Hunt, and V. H. Engelhard. Proteasome-independent Ag processing is responsible for a significant fraction of the peptides presented by many human class I MHC alleles. Submitted for publication.