The JI
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     
 

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow Request Permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Pelte, C.
Right arrow Articles by Kern, F.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Pelte, C.
Right arrow Articles by Kern, F.
The Journal of Immunology, 2004, 172: 6783-6789.
Copyright © 2004 by The American Association of Immunologists

Random Screening of Proteins for HLA-A*0201-Binding Nine-Amino Acid Peptides Is Not Sufficient for Identifying CD8 T Cell Epitopes Recognized in the Context of HLA-A*02011

Christian Pelte*, Georgy Cherepnev*, Yanjun Wang*, Constanze Schoenemann{dagger}, Hans-Dieter Volk* and Florian Kern2,*

* Institut für Medizinische Immunologie der Charité, Humboldt-Universität zu Berlin (Charité), Campus Charité Mitte, and {dagger} Institut für Transfusionsmedizin und Immunhämatologie, Campus Charité Virchow Klinikum, Berlin, Germany


   Abstract
Top
 Abstract
Introduction
 Materials and Methods
 Results
 Discussion
 References
 
HLA-A2 is the most frequent HLA molecule in Caucasians with HLA-A*0201 representing the most frequent allele; it was also the first human HLA allele for which peptide binding prediction was developed. The Bioinformatics and Molecular Analysis Section of the National Institutes of Health (BIMAS) and the University of Tübingen (Syfpeithi) provide the most popular prediction algorithms of peptide/MHC interaction on the World Wide Web. To test these predictions, HLA-A*0201-binding nine-amino acid peptides were searched by both algorithms in 19 structural CMV proteins. According to Syfpeithi, the top 2% of predicted peptides should contain the naturally presented epitopes in 80% of predictions (www.syfpeithi.de). Because of the high number of predicted peptides, the analysis was limited to 10 randomly chosen proteins. The top 2% of peptides predicted by both algorithms were synthesized corresponding to 261 peptides in total. PBMC from 10 HLA-A*0201-positive and CMV-seropositive healthy blood donors were tested by ex vivo stimulation with all 261 peptides using crossover peptide pools. IFN-{gamma} production in T cells measured by CFC was used as readout. However, only one peptide was found to be stimulating in one single donor. As a result of this work, we report a potential new T cell target protein, one previously unknown CD8-T cell-stimulating peptide, and an extensive list of CMV-derived potentially strong HLA-A*0201-binding peptides that are not recognized by T cells of HLA-A*0201-positive CMV-seropositive donors. We conclude that MHC/peptide binding predictions are helpful for locating epitopes in known target proteins but not necessarily for screening epitopes in proteins not known to be T cell targets.


   Introduction
Top
 Abstract
 Introduction
Materials and Methods
 Results
 Discussion
 References
 
T cell epitope prediction is becoming increasingly fashionable as pathogen amino acid sequences are available in huge databases and computers allow the processing of large datasets in a short time. Ever since the quest for a vaccination against HIV infection started mobilizing a lot of human and financial resources, the World Wide Web has hosted an increasing number of sites providing epitope prediction, information on epitopes, and other information relevant to this subject. The only alternative to the use of predictions is empirical work. If conducted on a large number of organisms, empirical approaches mean a large workload and require efficient technology. In this regard, the use of rapid cytokine induction as readout (1, 2) is to be favored over conventional methods, such as proliferation or cytotoxicity studies. Using cytokine flow cytometry (CFC)3 with a short ex vivo incubation time, the work may be conducted over 2 or 3 consecutive days on the same material using protein-spanning pools of peptides for identifying protein targets on day 1 before proceeding to epitope identification on day 2 or 3 (2, 3). Such an approach is useful for large organisms, particularly if protein targets are not known. However, for organisms with hundreds or thousands of proteins, peptide synthesis even with modern highly parallel approaches (4) and testing is bound to be a major limitation. At this point, epitope prediction (5, 6) may be useful for lack of a feasible empirical approach. For testing such a strategy, we processed 19 structural CMV proteins by two prediction algorithms available on the World Wide Web, referred to as "Syfpeithi" (University of Tübingen; www.syfpeithi.de) (5) and "BIMAS" (Bioinformatics and Molecular Analysis Section of the National Institutes of Health; bimas.dcrt.nih.gov/molbio/hla_bind) (6). We limited our search to HLA-A*0201-presented peptides, because it is the most frequent HLA allele in the Caucasian population. Although Syfpeithi provides scores based on the presence of certain amino acids in certain positions along the MHC-binding groove, the BIMAS score, at least with respect to HLA-A*0201, is based on the calculation of the theoretical half-life of the MHC-I/{beta}2-microglobulin/peptide complex, which is a measure of peptide-binding affinity. The philosophy of both algorithms is essentially that the higher the binding affinity of a peptide to the MHC, the higher the likelihood that this peptide represents an epitope. According to Syfpeithi, naturally presented epitopes should be among the top-scoring 2% of all peptides predicted in 80% of predictions. Additional information can be obtained at the indicated Web sites and the quoted literature (5, 6). We decided to follow both predictions for the synthesis of a total of 261 peptides from the selected proteins, representing the top 2% of predicted peptides for each method. All peptides were tested in 10 donors who were all human CMV seropositive (IgG) by serology and HLA-A*0201 positive (DNA typing).


   Materials and Methods
Top
 Abstract
 Introduction
 Materials and Methods
Results
 Discussion
 References
 
Donors

All donors were healthy young men and women who were regular blood donors at the local blood bank (Blood Bank, Institute for Transfusion Medicine, Charité Campus Mitte, Berlin, Germany) or students recruited by the laboratory (Institute for Medical Immunology, Charité Campus Mitte, Berlin, Germany). Blood was drawn following informed consent. The study was approved by the local ethics committee.

Cell preparation and stimulation procedure

PBMC were prepared from 50 to 80 ml of freshly drawn citrated venous blood by standard Ficoll-Paque (Pharmacia, Uppsala, Sweden) density gradient centrifugation. Cells were then washed twice with sterile PBS and resuspended in "supplemented medium", i.e., RPMI 1640 medium (Biochrom, Berlin, Germany) containing 2 mmol/l L-glutamine (Biochrom), 10% (v/v) heat-inactivated FCS (Biochrom), and 100 IU penicillin/streptomycin (Biochrom) at a concentration of 1.0–2.0 x 106 cells/ml. Four microliters of individual peptide or peptide pool solution were added to 100 µl of supplemented medium placed in Falcon 2052 tubes (BD Biosciences, Heidelberg, Germany) before adding 400 µl of the cell suspension. The final concentration of each peptide (whether added individually or in a pool) was 1 µg/ml. An equivalent amount of DMSO was added to unstimulated samples. The final DMSO concentration in all assays including unstimulated controls was below 0.1% (v/v). All samples were placed in a standard incubator at a 5° slant (37°C, humidified 5% CO2 atmosphere). To prevent cytokine secretion from activated T cells, 2 µl of brefeldin A stock solution (5 mg/ml in DMSO) were dissolved in 500 µl of warm supplemented medium and added to the culture after 2 h (the final concentration of brefeldin A was 10 µg/ml). After an additional 4 h, the cells were washed with 3 ml of ice-cold PBS (430 x g, 8 min, 4°C) and resuspended in 3 ml of PBS containing 2 mM EDTA (Sigma-Aldrich, Taufkirchen, Germany), incubated for 10 min at 37°C (water bath), spun down (430 x g, 8 min, 4°C), and vortexed for 30 s. Cells were then washed with PBS containing 0.5% BSA (Serva, Heidelberg, Germany) and 0.1% sodium-azide (Sigma-Aldrich) (washing buffer) and the supernatant was removed by decanting. Fixation/permeabilization was performed in two steps using commercial lysing solution and Perm II solution (both from BD Biosciences). After careful resuspension of the pellets, 1 ml of lysing solution was added and the tubes were incubated for 10 min at room temperature in the dark. Cells were spun down (430 x g, 8 min, 4°C), the supernatant was removed and the pellet was resuspended. Then, 0.5 ml of Perm II solution was added to each tube. Following an additional wash step, combined surface and intracellular Ab staining was performed for 30 min on melting ice in the dark. Staining was performed in a final volume of 100 µl. All Abs were titered beforehand to achieve optimum staining results. Following a final wash step, the samples were analyzed on a flow cytometer.

Antibodies

The following mAbs were purchased from BD Biosciences: FITC-conjugated anti-IFN-{gamma}, PE-conjugated anti-CD69, PerCP-conjugated anti-CD3, and allophycocyanin-conjugated anti-CD8. PE-conjugated anti-CD4 was purchased from Coulter-Immunotech (Marseille, France).

Prediction of HLA-A*0201 presented immunogenic nine-amino acid peptides

All protein sequences were submitted to analysis by computerized HLA-binding prediction based on the freely accessible online databases, Syfpeithi (http://syfpeithi.bmi-heidelberg.com) and the "HLA-peptide binding prediction" site supplied by BIMAS (http://bimas.dcrt.nih.gov/molbio/hla_bind). Both programs provide peptide sequences that are likely to be presented by the selected HLA molecules along with a ranking or score. Although Syfpeithi provides scores based on the presence of certain amino acids in certain positions of the class I MHC-binding groove (anchor motifs), BIMAS supplies a ranking according to the estimated half-life of the peptide/MHC {alpha}-chain/{beta}2-microglobulin complex. The results of such an analysis are shown in Table I. The top 2% of peptides from each analysis were chosen for each protein.


View this table:
[in this window]
[in a new window]
  		Table I. Peptides selected from the UL75 amino acid sequence

 
Peptides

All peptides were purchased from NMI (Tübingen, Germany). Stock solutions of single peptides (80 mg/ml) were produced by dissolving freeze-dried peptides in DMSO (Pierce, Rockford, IL), and kept at –70°C. Working solution was prepared from stocks by dilution with DMSO. The final concentration of each peptide in all pools was 0.25 mg/ml. Peptide pools were frozen in aliquots of 4 µl at –70°C. The following proteins/peptides were chosen (for each protein, the peptides predicted by Syfpeithi are listed first. Additional peptides predicted by BIMAS, but not Syfpeithi, are listed second): 1) UL25 (SwissProt P16761)/Syfpeithi: ALFDFLRVV, LIYVQLPSL, ALYASTPAL, ALTSLVDPV, YLQHQLQSI, SLREDYAQL, SLRRLDEEL, ALYLYRQNL, LVDPVLNNV, AAALALHFL, RLTEVYQTL, TLNDLAFLV, LLACPDRPI, KLITRDICV. UL25/BIMAS (predicted in addition): TLLKIFSQV, VLFYTAHYT, QLSDVIYWA, SELCYLIYV, AVIINYYYV, FLVGVELMI. 2) UL46 (SwissProt P16783)/Syfpeithi: LLHRRLETL, HLYISLYLL, YISLYLLGI, CLLESVYTA, LMIPKDMYL, FLVHKELKL, RLDNFSVEL. UL46/BIMAS (predicted in addition): ELGDFREFV, LLGIRASNV, QALTARFFV. 3) UL48 (SwissProt P16785)/Syfpeithi: FLEEALAQI, SLGAFTRGL, CLGVHVADL, AIMEEGARL, YVLVTVNSL, GLHQVLMVL, VLTGFGVQL, ALFLYMLDV, FLTDRLQQL, GVYEHLAGL, VIREQLSYL, DLLSALQQL, LLQRLGQTL, IMRDRLRRL, KLFCYLVST, SLQEFCVLI, KLVQQLCQV, SVSELVQTL, ALGSWLFGI, WLFGIPVCL, HLGGTSAVL, SLARAVVFT, ALIAWVEEM, KMLELQMDL, RLFYDLRDL, PLLQEILQL, LVQEQLAQL, DLQRVVDAV, ALQQLSKKL, SLNHLDQIL, LLGKATQQL, RLGQTLGDL, MLQCLWLEL, WLQAQIRRL, SALEILRLV, TAAPSLPPL, SLPPLPDPI, ALGRYQQLV, HILNGFLPV, QIHDLLHVI, RLINAVLSM, GVGPQVTEL, LLDKFNQDL, LIHETQQAL, ALSDEADLL, ALMEKPLET, TLRELLGKA, MLAFLEEAL, TLATEYFAL, DLRTRFAEL, GLNALERHV, SLVYRLEDI, YRLEDIPSV, ALRDDGRPL, QLPTGEPVL, DLLKGQRIL, LLKGQRILV, ILRLVRLDL. UL48/BIMAS (predicted in addition): WLYAMATLV, KLFWENKLV, LMVLTGFGV, VLSMFHTLV, RLSLQEFCV, WMLAFLEEA, KLWLYAMAT, GLLDVFIQT, ALYPEYIYT, SQYDLRPYL, LLHVIETLV, FLYMLDVAT, GIFDFLRYA, LVLDEAFPT, IMFLHALHL, LLQDTWTET, QLPEVQQRL, TLLDQQADA, KVQELLQRL, KVQQDLQRV, RLYEEEEET. 4) UL56 (SwissProt P16724)/Syfpeithi: QLLELLRRL, HLAGVLSAL, LLGQQIYEL, ALHEDTALL, SLNKRLQGL, RIIDLITSL, NLLQKLCVV, LLTLYLEML, YLEMLLKAL, DIPERIYSL, LLLRHEGSL, GILAPEAGL, SIACLRDEL, GLLCNHIAV, AMSGTGTTL, RLTSKLIDV, NLIHKSLPV, HVRELVLSV. UL56/BIMAS (predicted in addition): FTYDEHLYV, TVYPSEWMV, TLNDIERFL, ALLDRALMA, RMFVGSVFA, MELECLKYC, LCLDFISKL, WMHVRELVL. 5) UL75 (SwissProt P12824), see Table I. 6) UL80.a (SwissProt P16753), see Table II. 7) UL85 (SwissProt P16728)/Syfpeithi: LLIMGLFSL, KLTKLVAAV, LLRNMTLTL, LLHVPTHGL, ALFVMLRQL, KLVAAVVPI, LARENLLTL, LLTLGQWEL, LIDMKTACL, MLLVKCQEL, VMLRQLDDL. UL85/BIMAS (predicted in addition): YQLGLHQFV. 8) UL86 (SwissProt P16729)/Syfpeithi: SALELLPKV, TILDKILNV, LIHSFLQTL, TLAAMLYKI, VIMENLRRV, GLGLNLKTL, GLNLKTLLV, NLKTLLVDL, LLMPAATAV, MLTELVEDV, TLIEYSLPV, AMLYKISPV, ILSTTTLAL, AMHTVLRAL, KMLATLFLL, YLLNRDRAV, VLHEPAPCL, LLFHYRNLV, RLVTRISAL, ILTPVTMDV, SLTPLSMDV, NLVAVLRLV, PLSAYVNAL, AVGEMLTEL, VLYNGCCVV, QLIENPCRL, TLALMETKL, FLTSGLAAA, GLAAAAHAI, YVLKFLTRL, ALPTTAYLL, ALKTLCHPV, TLRLFKTTV, SIAAQRQAV, LTQEALPIL. UL86/BIMAS (predicted in addition): WTLQKVFYL, ELFLAVQFV, FLLIRTFVA, FVLSPENAV, YLPEDRGYT, KVLEVRAPL, YLLRAKDCI, FLLMPAATA, CLSDVLYNT, VMQIIMSLL, NMFHTRQLL, FVVTDFYKV. 9) UL100 (SwissProt P16733)/Syfpeithi: TLIVNLVEV, FLAVTIYYL, AIISIIYFL, GLCGLIYPI, SLTAFLFIL, SMIAFMAAV, ILSMDTFQL, VLAVFVVYA. UL100/BIMAS (predicted in addition): GMLFFIWAM, FMAAVHFFC, IVFMVLTFV, FMQLVFLAV, FMVLTFVNV. 10) UL115 (SwissProt P16832)/Syfpeithi: LLLPIVSSV, SLFNVVVAI, QLSAPTGSL, QLRALLTLL, VLLWCCLLL, HLDKYYAGL. UL115/BIMAS (predicted in addition): LLYNNPDQL, LVPPSLFNV, TLFYGLYNA, LLDDAFLDT.


View this table:
[in this window]
[in a new window]
  		Table II. MHC binding and proteasome cleavage predictions for peptides selected from CMV AD169, UL80.a

 
Setup of peptide pools and identification of single stimulating peptides

Peptide pools were compiled in such a way that each peptide was contained in exactly two pools. All peptides contained in a "vertical" pool (table column) were always from only one protein (see Table III). "Horizontal" pools contained peptides derived from various proteins and were designed to contain approximately the same number of peptides each. As previously described, positive stimulation results with these vertical and horizontal pools would identify intersections representing individual stimulating peptides. These were to be tested individually using the remaining PBMC suspension from day 1 (2, 7, 8).


View this table:
[in this window]
[in a new window]
  		Table III. Selected open reading frames and setup of peptide pools

 
Stabilization of HLA-A*0201 on class-I-HLA-transfected B x T hybrid cell lines (T2 lines)

MHC stabilization was performed according to the protocol published by Cerundolo et al. (9). Briefly, T2 cells (a gift from P. Walden, Department of Dermatology, Charité, Berlin, Germany) were incubated in RPMI 1640 medium containing 10% FSC and 2 mM glutamine with 100 µg (corresponding to 1.25 µl of peptide solution at 80 µg/ml in DMSO) of either the peptide to be tested, a positive control peptide (NLVPMVATV), or no peptide (only DMSO added) in a final volume of 1 ml at 37°C for 18 h. Cells were stained with a PE-conjugated anti-HLA-A2-specific Ab from Serotec (clone BB7.2; Düsseldorf, Germany) using 10 µl of Ab per 100 µl of staining volume for 30 min at room temperature. Cells were washed once before analysis on a BD Biosciences FACSCalibur flow cytometer.

Flow cytometric analysis

Data acquisition was conducted on a BD Biosciences FACSCalibur flow-cytometer using CellQuest software. A minimum of 250,000 lymphocytes (live count gate) were acquired for each sample. Twenty-five-thousand events were collected for the T2-stabilization assay. For data analysis, IFN-{gamma}+/CD3+/CD8+ events were gated and the percentages of the reference population (total CD8+ T cells) were determined.


   Results
Top
 Abstract
 Introduction
 Materials and Methods
 Results
Discussion
 References
 
All 261 predicted and synthesized peptides were tested with respect to their ability to induce IFN-{gamma} production in PBMC of all 10 donors. Table III shows the selected proteins (column heads), the number of peptides selected from each protein (bottom) and the setup of the peptide pools. Table I shows the peptides selected from UL75. Fifteen peptides (representing the top 2% of 735 possible nine-amino acid peptides) were selected using each algorithm, duplicates were eliminated. The remaining 23 peptides were used for testing. A complete list of all other tested peptides is provided in Materials and Methods. Intracellular IFN-{gamma} was detected using a previously published standard CFC-based assay (1). In our experience, frequencies down to ~1 or 2/10,000 cells can be clearly distinguished from background (8).

For each experiment, 38 single tubes were run, including 36 tubes with pools of the candidate peptides (see Table III), a positive and a negative control tube. We decided to use the known HLA-A*0201-presented peptide, NLVPMVATV (10), as "positive control", however, because some HLA-A*0201 CMV-responsive donors did not respond to this peptide in the past, it was not a positive control in the strict sense. In this study, however, all tested donors had a response to NLVPMVATV, indicating that negative responses to peptide pools were not due to the experimental setup. An unstimulated sample was used as negative control.

Only 1 of 10 donors had a detectable T cell response to any of the pools (Fig. 1A). He responded to pools 14 and 25, identifying peptide 172 as stimulating (peptide number in bold print in Table III). This response was confirmed using the individual peptide to stimulate stored PBMC from the same donor. The experiment was repeated several weeks later and the initial results were confirmed. The identified peptide belongs to the CMV capsid protein P40 (UL-80.a), and has the amino acid sequence GASPAVSSL (P4022–28). Unfortunately, none of the remaining donors had a response to any of the peptide pools.



View larger version (31K):
[in this window]
[in a new window]
  		FIGURE 1. T cell response and MHC stabilization induced by the peptides, GASPAVSSL and NLVPMVATV. A, PBMC from an HLA-A*0201-positive donor (donor 7 in Table IV) were incubated for 6 h under standard culture conditions with the indicated peptides or peptide pools at a final concentration of 1 µg/ml per peptide. Brefeldin A was added after 2 h. Diagrams are gated on CD3+ events. Numbers indicate percentages of IFN-{gamma}+ events with respect to the CD8+CD3+ T cell subset. CD8low/IFN-{gamma}+ events observed in some panels may include CD4+ T cells stimulated by nine-amino acid peptides. Axes show log fluorescence intensities. B, The peptides GASPAVSSL and NLVPMVATV were compared in regard to their ability to stabilize HLA-A*0201 expression on T2 cells. Both peptides were incubated with T2 cells for 18 h at 37°C. The diagram shows that while HLA-A*0201 expression was clearly increased by NLVPMVATV, GASPAVSSL was unable to increase expression over background level (control).

 
Although all the predicted peptides were supposed to be presented by HLA-A*0201, we felt it was rather unlikely that the stimulating peptide, GASPAVSSL, was in fact presented by this molecule. It is known that some CMV-positive HLA-A*0201-positive donors do not respond to the "positive control" peptide NLVPMVATV, however, a majority do (10 of 10 in this study). It was previously shown that the HLA-A*0201-presented HIVgag peptide SLYNTVATL is recognized in maybe half of HIV-positive patients possessing the said allele (11), however, there has been no report yet of a peptide definitely presented by HLA-A*0201 but recognized by as little as 1 in 10 donors with the right HLA type and an immune response against the respective Ag. Therefore, we examined whether GASPAVSL was able to stabilize the expression of HLA-A*0201 on T2 cells, which is a gauge of its ability to bind to this molecule. T2 cells were incubated with no peptide, NLVPMVATV (positive control) and GASPAVSSL. The histogram in Fig. 1B shows that GASPAVSSL was unable to stabilize HLA-A*0201 expression (geometric mean fluorescence intensity 10.87 and 10.30 for no peptide and GASPAVSSL, respectively), while NLVPMVATV clearly was (geometric mean fluorescence intensity 24.07). Because binding to HLA-A*0201 is prerequisite for being an epitope presented in this context, GASPAVSSL must be presented by a different HLA molecule. Table IV displays the other class I HLA alleles (A and B locus, the C locus was not determined) of the donors included in this study. Donor 7 responded to GASPAVSSL (underlined). He shared three more HLA molecules with two other donors (HLA-A*2404, HLA-B*40, and HLA-B*57). HLA-binding predictions for HLA-B40 and HLA-B57 were not available on the Syfpeithi server, the prediction score for GASPAVSSL for binding to HLA-A*2402 was low. BIMAS offers predictions for HLA-A24 (4.800) and HLA-B40 (2.000), but not HLA-B57. No definite conclusions about the presenting HLA molecule can be drawn from the available data, so that additional experiments need to be performed to identify the allomorph presenting GASPAVSSL.


View this table:
[in this window]
[in a new window]
  		Table IV. Donor HLA types (A and B loci)

 
Despite the fact that GASPAVSSL is not presented on HLA-A*0201, we were interested to learn whether the inclusion of results from proteasome cleavage prediction would have limited the choice of our peptides in such a way that GASPAVSSL would have been excluded. Table II displays the various outputs for the selected UL80.a-derived peptides generated by Syfpeithi and BIMAS (the two selection criteria) as well as Fragpredict (fragment cleavage probability), PaproC (C-terminal cut only, cleavage strength based on analyses for wild types I-III), and Netchop (C-terminal cleavage, probability). Based on these results, only one peptide would have been excluded if all peptides predicted to be cleaved by at least one of the algorithms would have been included. If used alone, each cleavage prediction program would have lead to the exclusion of two peptides. Such good correlations between different predictive algorithms do not always exist. We used Fragpredict to also analyze all other proteins and found that, for example, only 3 of 20 peptides predicted by Syfpeithi and BIMAS for UL25 (top 2%), would have been predicted to be cleaved even with a lower cleavage probability of 50%. This indicates that the results for UL80.a do not necessarily extend to other proteins. A comprehensive and comparative analysis of all available epitope, peptide-MHC binding, and proteasome cleavage predictions, however, is not within the scope of this work.


   Discussion
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
References
 
The goal of this study was to test a strategy for searching for HLA-A*0201-presented stimulating peptides in previously not examined proteins of the human CMV. So, indirectly the ability of MHC-peptide binding (BIMAS) and epitope predictions (Syfpeithi) to predict not only epitopes but also protein targets of the CD8 T cell response were tested. The result of this extensive work is a list of 260 peptides ~82% of which can be expected to bind very well to HLA-A*0201 (12) but none of which was recognized by the T cell response in 10 donors in this HLA context. However, the results with the "control" peptide, NLVPMVATV indicated clearly that all of them were able to respond to an HLA-A*0201-presented peptide from CMV. We identified one single stimulating peptide, GASPAVSSL (P4022–28), which, however, did not stabilize HLA-A*0201 expression on T2 cells and could therefore not have been presented by HLA-A*0201. This was in agreement with the low response rate to this peptide (1 of 10) in HLA-A*0201-positive donors. As an additional result, the CMV capsid protein P40, was identified as a potential target of the T cell response. Most importantly, however, there is a lesson to be learned from this study. Two MHC/peptide-binding prediction algorithms were put to the test as a sole means of identifying new epitopes and new target proteins at the same time. The prediction of epitopes is meant to save money and time by shortcutting the process of a step-by-step empirical analysis. With regard to HLA-A2 and its A*0201 allele, the predictions by Syfpeithi are based on a very large number of known ligands, and known stimulating HLA-A*0201-presented CMV-derived peptides get high ratings or scores. Moreover, the results published by Elkington et al. (12) indicate that Syfpeithi and BIMAS are good at predicting binding to HLA-A*0201. The CMV pp65-derived peptide, NLVPMVATVQ, for example, was first discovered to stimulate CD8 T cells in an HLA-A2 context by an empirical approach by Wills et al. (10) in 1996. The same epitope was then rediscovered using first an empirical approach to narrow down stimulating protein sections and then the BIMAS prediction algorithm to define the epitope, this time, a nine-amino acid peptide, NLVPMVATV (13). Ironically, using just the BIMAS prediction on pp65 it also comes up as no. 4 on the list. NLVPMVATV obtains a score of 159.970 and 30, by BIMAS and Syfpeithi, respectively. Interestingly, using Syfpeithi and BIMAS for processing pp65, three known HLA-A*0201-presented peptides, VLGPISGHV, (pp6514–22), MLNIPSINV (pp65120–128), and NLVPMVATV(pp65495–503) come up among the first 13 and 20 nine-amino acid peptides, respectively. The peptides selected for this study obtained a BIMAS score ranging from 6.086 to 12,279.298 and a Syfpeithi score ranging from 23 to 32. Nevertheless, none of the peptides seemed to represent an epitope in the HLA-A*0201 context. The reason for this may be that none of the two algorithms can account for factors outside MHC/peptide binding that may, however, be responsible for making a protein a T cell target. Apart from the fact that peptides must bind to MHC molecules to form epitopes, they must also be obtained by proteasome cleavage or other processing of the original protein sequence, and TCRs must exist that bind to these complexes with sufficient affinity for the T cells to be activated. Although programs for predicting TCR binding to MHC/peptide complexes are not in the public domain to date, there are several programs predicting proteasome cleavage. Proteasome cleavage prediction programs available on the Web include Fragpredict (www.mpiib-berlin.mpg.de/MAPPP/cleavage.html), PaproC (www.paproc.de), and Netchop (www.cbs.dtu.dk/services/NetChop). As would be expected, the lower the cleavage probability is chosen (possible with Fragpredict and Netchop), the higher the number of cleavage sites that these programs predict. Although Fragpredict supplies peptides with an associated cleavage probability, both Netchop and PaproC list C-terminal cleavage sites. The number of peptides/cleavage sites listed by these programs using normal default values for cleavage prediction is usually large, so that they are likely to be used only in combination with MHC-binding predictions rather than alone. However, in the present study, such predictions were not considered when choosing the peptides because we wanted to test the Syfpeithi and BIMAS algorithms in particular and with a view to predicting epitopes (and indirectly T cell target proteins). By deselecting peptides according to other predictions we would not have been able to do so. Also, we felt it would have given too much weight to cleavage predictions whose relevance may be questioned in view of the fact that not all peptides presented on class I MHC molecules necessarily result from proteasome processing (14, 15).

Nevertheless, we analyzed, in retrospect, whether the one stimulating peptide, GASPAVSSL, would have been omitted if only those peptides also predicted to result from proteasome cleavage had been included. Table II illustrates that this is not the case. In fact, if any peptide predicted to be processed by at least one of them had been included, only one peptide would have been deselected.

A combination of epitope prediction and cleavage prediction is an attractive option, however, relatively tedious if done by using all prediction programs separately. The MHC-I Antigenic Peptide Processing Prediction (MAPPP) website (www.mpiib-berlin.mpg.de/MAPPP/index.html) offers a combined program that selects peptides achieving certain minimum scores, for example by Fragpredict, Syfpeithi, and BIMAS. These minimum scores are set below the scores one would normally use if each program was used alone. The original scores are transformed and all selected peptides are then sorted by an overall scoring system. This way peptides with a good score in one and a reasonable score in the other algorithm are not automatically eliminated, increasing the chances of identifying epitopes. Unfortunately, we did not have the resources to repeat the study using the combined predictions by MAPPP.

In a recent publication, Elkington et al. (12) present an approach to searching new CMV epitopes similar to the one we have used in this study. First, potential MHC-presented peptides were predicted using Syfpeithi or BIMAS algorithms. Second, the selected peptides were tested in MHC-stabilization assays using T2 cells. Then, they performed ELISPOT analysis to see whether these peptides were able to induce IFN-{gamma} in PBMC from donors selected for CMV seropositivity and by HLA type. Finally, they performed cytotoxicity tests with some of the peptides on autologous donor B-LCL lines. They eventually present a list of >60 peptides, which were positive in either ELISPOT or cytotoxicity assay or both. Although their comprehensive list of new epitopes is rather impressive at first glance, there are clearly some problems with the interpretation of the data. For example, there are two peptides that are listed as HLA-A2-presented epitopes from CMV US2, LLVLFIVYV, and TLLVLFIVYV. Interestingly, both failed to stabilize HLA-A2. LLVLFIVYV gave weakly positive results in only two of eight HLA-A2-positive donors in ELISPOT, neither of the two, however, responded to TLLVLFIVYV. TLLVLFIVYV by contrast gave only one positive response in one additional donor of these eight. Cytotoxicity assays were not performed with these peptides. As a result, the conclusion that these are class I MHC-restricted and HLA-A2-presented peptides is based solely on the prediction algorithms and clearly stands in contrast to the result of MHC stabilization. Another example is the glycoprotein H (UL75). Interestingly, this protein was the only one that was examined by this group and ourselves. The peptides we tested included all the peptides that Elkington et al. (12) tested with respect to HLA-A*0201 and this protein. We found no response to any of these peptides in our donors. In contrast to this, Elkington et al. (12) found that the peptide, LIFGHLPRV (gH249–257), caused a very weak ELISPOT response in one of eight HLA-A2-positive (no alleles given) donors. This peptide stabilized HLA-A2, however, cytotoxicity tests were not done. Because only one individual responded, they concluded that this peptide represents a "subdominant" HLA-A2-associated epitope. Additionally they found two weak and two very weak ELISPOT responses in the same eight donors for the peptide, LMLLKNGTV. According to the provided data, however, HLA stabilization by this peptide was not tested and cytotoxicity testing also was not performed. The inference that it is HLA-A2 presented is based mainly on the predictions and some weak ELISPOT responses. One of the major drawbacks of using ELISPOT in such a study is the fact that CD4 and CD8 responses cannot be distinguished unless cells are presorted. Both CD4 and CD8 T cells produce IFN-{gamma} upon stimulation with peptides, and importantly, CD4 T cells can often be stimulated by nine-amino acid peptides. Such responses may be suboptimal yet detectable. In this situation, compared with ELISPOT, CFC has the advantage of being able to identify the responding T-lymphocyte lineage. This is one reason why CFC was used in this study. According to a recent systematic study, the sensitivities of the two methods for detecting IFN-{gamma} in CD8 T cells following peptide stimulation are in fact the same (16). Although we much appreciate the work by Elkington et al. (12), we are not convinced by the presented data that all responses were CD8 T cell responses. Also the list of presenting HLA molecules may be debated in part based on their own data. We are currently performing an empirical mapping (all possible nine-amino acid peptides) of several large CMV proteins including the ones we tested here, and we should hopefully be able to use this dataset to test prediction algorithms more extensively in the future. At this stage, it seems that these predictions work well with respect to predicting binding, however, some proteins may never get a chance to become T cell targets, making binding predictions irrelevant. Proteins that are coded for by open reading frames but not expressed, for example, cannot possibly be targets. Also, some proteins may be expressed but are protected from processing by the proteasome as described in the murine CMV model (17). It is tempting to argue that structural proteins of viruses, for example, are more likely to be targets than nonstructural proteins, or, that abundant proteins are more likely to be presented than rare ones, but truly we do not understand very well why some proteins are T cell targets but not others. Interestingly, in 1993, it was reported that somehow the pp65 protein prevents presentation of IE-1 peptides by mediating methylation of some of its amino acid side chains, explaining that IE-1 was not an important T cell target (18), which was in keeping with the current opinion at the time. Nevertheless, the murine CMV IE-1 had been described to be a T cell target in the mouse system several years earlier (19, 20). It was not until 1999 that IE-1 was rediscovered as a T cell target in humans (7). It is possible that IE-1 is a good T cell target, only, because during CMV reactivation, IE-1 is expressed before pp65. A very interesting review on the role of CMV proteins in immune surveillance/immune evasion was recently published (17). Supposing that the P40 protein identified in this study by a strong T cell response to GASPAVSSL really is a CD8 T cell target (cross-reactivity of the identified peptide cannot be ruled out at this point), there is nothing we know about this protein that would have let us guess so, except maybe that it is structural (21). In this study, we were clearly not looking for epitopes in known target proteins but, indirectly, for new target proteins as well, and apart from the MHC/peptide binding predictions there was nothing to predict that antigenic determinants from these proteins would in fact be recognized in the context of HLA-A*0201.


   Footnotes
 
1 This work was in part supported by the Deutsche Forschungsgemeinschaft (FOR 299). Back

2 Address correspondence and reprint requests to Dr. Florian Kern, Institut für Medizinische Immunologie der Charité, Charité Campus Mitte, Hochhaus Ebene 4, Schumannstrasse 21, 10117 Berlin, Germany. E-mail address: florian.kern{at}charite.de Back

3 Abbreviation used in this paper: CFC, cytokine flow cytometry. Back

Received for publication July 28, 2003. Accepted for publication March 26, 2004.


   References
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
 References
 

  1. Kern, F., I. P. Surel, C. Brock, B. Freistedt, H. Radtke, A. Scheffold, R. Blasczyk, P. Reinke, J. Schneider-Mergener, A. Radbruch, et al 1998. T-cell epitope mapping by flow cytometry. Nat. Med. 4:975.[Medline]
  2. Hoffmeister, B., F. Kiecker, L. Tesfa, H. D. Volk, L. J. Picker, F. Kern. 2003. Mapping T cell epitopes by flow cytometry. Methods 29:270.[Medline]
  3. Kern, F., N. Faulhaber, C. Frömmel, E. Khatamzas, S. Prösch, C. Schönemann, I. Kretzschmar, R. Volkmer-Engert, H.-D. Volk, P. Reinke. 2000. Analysis of CD8 T cell reactivity to cytomegalovirus using protein-spanning pools of overlapping pentadecapeptides. Eur. J. Immunol. 30:1676.[Medline]
  4. Kramer, A., U. Reineke, L. Dong, B. Hoffmann, U. Hoffmuller, D. Winkler, R. Volkmer-Engert, J. Schneider-Mergener. 1999. Spot synthesis: observations and optimizations. J. Pept. Res. 54:319.[Medline]
  5. Rammensee, H. G., J. Bachmann, S. Stefanovic. 1997. MHC Ligands and Binding Motifs Landes Bioscience, Georgetown.
  6. Parker, K. C., M. A. Bednarek, J. E. Coligan. 1994. Scheme for ranking potential HLA-A2 binding peptides based on independent binding of individual peptide side-chains. J. Immunol. 152:163.[Abstract]
  7. Kern, F., I. P. Surel, N. Faulhaber, C. Frommel, J. Schneider-Mergener, C. Schonemann, P. Reinke, H. D. Volk. 1999. Target structures of the CD8+-T-cell response to human cytomegalovirus: the 72-kilodalton major immediate-early protein revisited. J. Virol. 73:8179.[Abstract/Free Full Text]
  8. Kern, F., T. Bunde, N. Faulhaber, F. Kiecker, E. Khatamzas, I. M. Rudawski, A. Pruss, J. W. Gratama, R. Volkmer-Engert, R. Ewert, et al 2002. Cytomegalovirus (CMV) phosphoprotein 65 makes a large contribution to shaping the T cell repertoire in CMV-exposed individuals. J. Infect. Dis. 185:1709.[Medline]
  9. Cerundolo, V., J. Alexander, K. Anderson, C. Lamb, P. Cresswell, A. McMichael, F. Gotch, A. Townsend. 1990. Presentation of viral antigen controlled by a gene in the major histocompatibility complex. Nature 345:449.[Medline]
  10. Wills, M. R., A. J. Carmichael, K. Mynard, X. Jin, M. P. Weekes, B. Plachter, J. G. Sissons. 1996. The human cytotoxic T-lymphocyte (CTL) response to cytomegalovirus is dominated by structural protein pp65: frequency, specificity, and T- cell receptor usage of pp65-specific CTL. J. Virol. 70:7569.[Abstract]
  11. Betts, M. R., J. P. Casazza, B. A. Patterson, S. Waldrop, W. Trigona, T. M. Fu, F. Kern, L. J. Picker, R. A. Koup. 2000. Putative immunodominant human immunodeficiency virus-specific CD8+ T-cell responses cannot be predicted by major histocompatibility complex class I haplotype. J. Virol. 74:9144.[Abstract/Free Full Text]
  12. Elkington, R., S. Walker, T. Crough, M. Menzies, J. Tellam, M. Bharadwaj, R. Khanna. 2003. Ex vivo profiling of CD8+-T-cell responses to human cytomegalovirus reveals broad and multispecific reactivities in healthy virus carriers. J. Virol. 77:5226.[Abstract/Free Full Text]
  13. Diamond, D. J., J. York, J. Y. Sun, C. L. Wright, S. J. Forman. 1997. Development of a candidate HLA A*0201 restricted peptide-based vaccine against human cytomegalovirus infection. Blood 90:1751.[Abstract/Free Full Text]
  14. Nakagawa, Y., T. Takeshita, J. A. Berzofsky, H. Takahashi. 2000. Analysis of the mechanism for extracellular processing in the presentation of human immunodeficiency virus-1 envelope protein-derived peptide to epitope-specific cytotoxic T lymphocytes. Immunology 101:76.[Medline]
  15. Sherman, L. A., T. A. Burke, J. A. Biggs. 1992. Extracellular processing of peptide antigens that bind class I major histocompatibility molecules. J. Exp. Med. 175:1221.[Abstract/Free Full Text]
  16. Sun, Y., E. Iglesias, A. Samri, G. Kamkamidze, T. Decoville, G. Carcelain, B. Autran. 2003. A systematic comparison of methods to measure HIV-1 specific CD8 T cells. J. Immunol. Methods 272:23.[Medline]
  17. Reddehase, M. J.. 2002. Antigens and immunoevasins: opponents in cytomegalovirus immune surveillance. Nat. Rev. Immunol. 2:831.[Medline]
  18. Gilbert, M. J., S. R. Riddell, C. R. Li, P. D. Greenberg. 1993. Selective interference with class I major histocompatibility complex presentation of the major immediate-early protein following infection with human cytomegalovirus. J. Virol. 67:3461.[Abstract/Free Full Text]
  19. Del Val, M., H. Volkmer, J. B. Rothbard, S. Jonjic, M. Messerle, J. Schickedanz, M. J. Reddehase, U. H. Koszinowski. 1988. Molecular basis for cytolytic T-lymphocyte recognition of the murine cytomegalovirus immediate-early protein pp89. J. Virol. 62:3965.[Abstract/Free Full Text]
  20. Reddehase, M. J., J. B. Rothbard, U. H. Koszinowski. 1989. A pentapeptide as minimal antigenic determinant for MHC class I-restricted T lymphocytes. Nature 337:651.[Medline]
  21. Chee, M. S., A. T. Bankier, S. Beck, R. Bohni, C. M. Brown, R. Cerny, T. Horsnell, C. A. Hutchison, 3rd, T. Kouzarides, J. A. Martignetti, et al 1990. Analysis of the protein-coding content of the sequence of human cytomegalovirus strain AD169. Curr. Top. Microbiol. Immunol. 154:125.[Medline]



This article has been cited by other articles:


Home page
 J. Immunol.Home page
A. N. Hasan, W. J. Kollen, D. Trivedi, A. Selvakumar, B. Dupont, M. Sadelain, and R. J. O'Reilly
A Panel of Artificial APCs Expressing Prevalent HLA Alleles Permits Generation of Cytotoxic T Cells Specific for Both Dominant and Subdominant Viral Epitopes for Adoptive Therapy
J. Immunol., August 15, 2009; 183(4): 2837 - 2850.
[Abstract] [Full Text] [PDF]

Home page
 J. Immunol.Home page
A. A. Chentoufi, X. Zhang, K. Lamberth, G. Dasgupta, I. Bettahi, A. Nguyen, M. Wu, X. Zhu, A. Mohebbi, S. Buus, et al.
HLA-A*0201-Restricted CD8+ Cytotoxic T Lymphocyte Epitopes Identified from Herpes Simplex Virus Glycoprotein D
J. Immunol., January 1, 2008; 180(1): 426 - 437.
[Abstract] [Full Text] [PDF]

Home page
 BloodHome page
M. Wolfl, J. Kuball, W. Y. Ho, H. Nguyen, T. J. Manley, M. Bleakley, and P. D. Greenberg
Activation-induced expression of CD137 permits detection, isolation, and expansion of the full repertoire of CD8+ T cells responding to antigen without requiring knowledge of epitope specificities
Blood, July 1, 2007; 110(1): 201 - 210.
[Abstract] [Full Text] [PDF]

Home page
 DiabetesHome page
N. E. Standifer, Q. Ouyang, C. Panagiotopoulos, C. B. Verchere, R. Tan, C. J. Greenbaum, C. Pihoker, and G. T. Nepom
Identification of Novel HLA-A*0201-Restricted Epitopes in Recent-Onset Type 1 Diabetic Subjects and Antibody-Positive Relatives
Diabetes, November 1, 2006; 55(11): 3061 - 3067.
[Abstract] [Full Text] [PDF]

Home page
 DiabetesHome page
Q. Ouyang, N. E. Standifer, H. Qin, P. Gottlieb, C. B. Verchere, G. T. Nepom, R. Tan, and C. Panagiotopoulos
Recognition of HLA Class I-Restricted {beta}-Cell Epitopes in Type 1 Diabetes
Diabetes, November 1, 2006; 55(11): 3068 - 3074.
[Abstract] [Full Text] [PDF]

Home page
 Mol. Cell. ProteomicsHome page
H. D. Meiring, E. C. Soethout, M. C. M. Poelen, D. Mooibroek, R. Hoogerbrugge, H. Timmermans, C. J. Boog, A. J. R. Heck, A. P. J. M. de Jong, and C. A. C. M. van Els
Stable Isotope Tagging of Epitopes: A Highly Selective Strategy For The Identification Of Major Histocompatibility Complex Class I-Associated Peptides Induced Upon Viral Infection
Mol. Cell. Proteomics, May 1, 2006; 5(5): 902 - 913.
[Abstract] [Full Text] [PDF]

Home page
 J. Immunol.Home page
R. Khanna
Predictive Algorithms and T Cell Epitope Mapping
J. Immunol., September 1, 2004; 173(5): 2895 - 2895.
[Full Text] [PDF]

Home page
 J. Immunol.Home page
F. Kern
The Authors Respond
J. Immunol., September 1, 2004; 173(5): 2895 - 2896.
[Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow Request Permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Pelte, C.
Right arrow Articles by Kern, F.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Pelte, C.
Right arrow Articles by Kern, F.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS