ava i l ab l e a t www.sc i enced i r ec t . com

www.e l sev i e r . com/ l oca te /yc l im

Multiplex mapping of CD4 T cell epitopesusing class II tetramers

Junbao Yang a, Eddie A. James a, Laurie Huston a, Nancy A. Danke a,Andrew W. Liu a, William W. Kwok a,b,*

a Benaroya Research Institute at Virginia Mason, 1201 Ninth Ave., Seattle, WA 98101, USAb Department of Immunology, University of Washington, 1959 NE Pacific Street, Seattle, WA 98195-7650, USA

Received 9 September 2005; accepted with revision 22 March 2006Available online 4 May 2006

1521-6616/$ — see front matter D 20doi:10.1016/j.clim.2006.03.008

* Corresponding author. Benaroya RMason, 1201 Ninth Ave., Seattle, WA7638.

E-mail address: bkwok@benaroyare

KEYWORDST cell epitope;Tetramer-guidedepitope mapping;Hemagglutinin;Matrix protein

Abstract With the advent of class II tetramer technology, a tetramer-guided epitope mapping(TGEM) technique was developed for the identification of CD4+ Tcell epitopes. This allowed thedirect identification of epitopes recognized by the responding T cells, which were restricted tothe single MHC allele of interest. However, as each individual carries multiple class II alleles, itwould be advantageous to design an approach to identify CD4+ epitopes presented by differentclass II alleles at the same time. In the present study, a multiplex TGEM approach was developedto identify antigenic epitopes presented by multiple HLA class II alleles simultaneously. In thisnew approach, CD4+ T cells were stained with multiple sets of MHC class II tetramers—eachlabeled with a unique fluorescent label. Using this multiplex approach, novel epitopes frominfluenza antigens hemagglutinin and matrix protein presented by multiple class II alleles wereidentified in a single experimental setting.D 2006 Elsevier Inc. All rights reserved.

Introduction

Understanding T cell epitope recognition is important as abasic immunological question and for practical applicationssuch as vaccine development and immune modulation in thecontexts of allergy and autoimmunity. It is well known thatthe selection of epitopes from an antigenic protein is

06 Elsevier Inc. All rights reserv

esearch Institute at Virginia98101, USA. Fax: +1 206 223

search.org (W.W. Kwok).

dependent on antigen processing, peptide binding to theMHC, and interaction of the MHC/peptide complex with thespectrum of available TCR. In this way, the myriad ofpossible peptide patterns is narrowed to a much smallernumber of dominant epitopes. A variety of approaches havebeen developed for identification of T cell epitopes within agiven antigenic protein. These approaches are summarizedin Table 1.

Each of these methods is based either on the ability ofantigenic peptides to bind MHC or the ability of antigenicpeptides to elicit a measurable functional response. Whileeach of these approaches has facilitated epitope identifica-tion, some limitations remain. The first three approaches —

Clinical Immunology (2006) 120, 21—32

ed.

Table 1 Summary of methods for epitope identification

Method Limitation Reference

ELISpot assay Difficult to determine HLA restriction [1,2]Intracellular cytokine staining Difficult to determine HLA restriction [3,4]CFSE flow cytometry assay Difficult to determine HLA restriction [4]MHC-binding prediction algorithm Lack of accuracy and relevance as T cell epitope [5,6]Peptide elution analysis by massspectrometry/liquid chromatography(MS-HPLC)

Relevance as T cell epitope [7]

Combinatorial peptide library Natural epitope remains unknown [8]HLA transgenic mice Cost and time [10]

J. Yang et al.22

CFSE assays, ELISpot (enzyme-linked immunosorbent spotassay), and Intracellular cytokine staining — are sensitiveand capable of identifying epitopes without any priorknowledge of the HLA genotypes of the subject [1—4].However, subsequent determination of the HLA restrictionfor each specific epitope is difficult and time consuming.Computer-assisted prediction of MHC peptide binding is auseful approach to identify possible epitopes [5,6]. Howev-er, the accuracy of these predictions must be verifiedempirically. In addition, peptide binding to the MHC doesnot assure the presence of T cells specific for the MHC/peptide complex. Similarly, the elution of peptides fromantigen presenting cells and subsequent analysis by chro-matography and mass spectrometry can yield importantinformation about naturally processed epitopes from a givenantigen [7]. But again, the resulting epitopes must beinterrogated by a secondary method to measure the abilityof the epitope to elicit a T cell response. The use ofcombinatorial peptide libraries can generate substantialinformation about the reactivity to various peptide patterns[8]. However, there is a disconnection that must be bridgedbetween the reactive mimotopes and their biologicalcounterparts. In many cases, determination of the bnaturalQepitope that corresponds to the synthetic mimotope isdifficult or impossible [9]. HLA transgenic mice have beenused to map various T cell epitopes [10], but the time andexpenses associated with maintaining a mouse colony andgenerating transgenic mice with the desired HLA genotypeshave limited the widespread use of this approach.

With the advent of HLA class II tetramers, it is nowpossible to detect antigen-specific CD4+ T cells by flowcytometry. Our laboratory has developed a method forproducing functional tetramers without using a linkedpeptide [11]. This allows the production of numerouspeptide/MHC combinations by loading the MHC withexogenous peptide. More importantly, the MHC can beloaded with mixtures (or pools) of peptide, facilitating thesimultaneous screening of multiple peptides. Based onthese innovations, the tetramer guided epitope mapping(TGEM) method was developed to identify class IIrestricted T cell epitopes [12,13]. With this approach, Tcell epitopes restricted by one class II allele of interestwere identified in a single experimental setting within 3weeks. Following this and other successes [12—15], wereasoned that as each individual carries multiple allelesfor HLA-DR, DQ and DP, a large amount of epitopeinformation could be gleaned from each individual parti-cularly if multiple alleles could be tested using a single

sample. The goal of this present study, then, was toextend TGEM to allow the mapping of antigenic epitopesfor multiple class II alleles simultaneously. Our resultsdemonstrated that multiplex TGEM can be used to identifyT cell epitopes for as many as four class II allelessimultaneously—by using a distinct fluorescent marker tolabel tetramers for each allele. These experimentsdemonstrated that TGEM provides a high throughputapproach for T cell epitope identification.

Materials and methods

Blood samples

All donors were healthy subjects recently immunized withthe trivalent type A and B influenza virus bFluzoneQ vaccine(Aventis, Bridgewater, NJ). PBMC were isolated from hepa-rinized venous blood by Ficoll gradient centrifugation. HLAclass II typing of donors was performed by reverse dot blothybridization as described [16].

Chromophore-labeled antibodies and streptavidin

FITC-labeled anti-human CD4 antibody, PerCP-Cy5.5-labeledstreptavidin, and PerCP-labeled streptavidin were pur-chased from BD Biosciences (San Jose, CA). PE-labeledstreptavidin, FITC-labeled streptavidin, Alexa 488-labeledstreptavidin, and APC-labeled streptavidin were purchasedfrom Molecular Probes (Eugene, OR) and Biosource Interna-tional (Camarillo, CA).

Peptide panel for influenza epitope mapping

To allow a thorough study of influenza epitopes, a panelof 41 overlapping hemagglutinin (HA) peptides (derivedfrom HA1 subunit of influenza A virus A/New Caledonia/20/99 (H1N1)), designated as p1 to p41 (Table 2) and 30overlapping matrix protein (MP) peptides (derived fromMP of influenza A virus A/Puerto Rico/8/34 (H1N1)),designated as p42 to p71 (Table 3), were designed.These peptides, each 20 amino acids in length, coverthe entire sequence of Influenza HA1 and MP proteinswith a 12-amino-acid overlap between consecutive pep-tides. The peptides were synthesized on polyethylenepins using 9-fluorenylmethoxycarbonyl chemistry byMimotopes (Clayton, Australia). Individual peptides weredissolved in DMSO to achieve the appropriate peptide

Table 2 Overlapping peptides spanning the influenza A hemagglutinin (HA) protein

Peptide pools Peptide name Residues Peptide sequence MHC restriction

HA pool #1 P1 1—20 MKAKLLVLLCTFTATYADTIP2 9—28 LCTFTATYADTICIGYHANNP3 17—36 ADTICIGYHANNSTDTVDTVP4 25—44 HANNSTDTVDTVLEKNVTVTP5 33—52 VDTVLEKNVTVTHSVNLLED

HA pool #2 P6 41—60 VTVTHSVNLLEDSHNGKLCLP7 49—68 LLEDSHNGKLCLLKGIAPLQP8 57—76 KLCLLKGIAPLQLGNCSVAGP9 65—84 APLQLGNCSVAGWILGNPECP10 73—92 SVAGWILGNPECELLISKES

HA pool #3 P11 81—100 NPECELLISKESWSYIVETPP12 89—108 SKESWSYIVETPNPENGTCYP13 97—116 VETPNPENGTCYPGYFADYEP14 105—124 GTCYPGYFADYEELREQLSSP15a 113—132 ADYEELREQLSSVSSFERFE DRB5

HA pool #4 P16 121—140 QLSSVSSFERFEIFPKESSW DRB5

P17 129—148 ERFEIFPKESSWPNHTVTGVP18 137—156 ESSWPNHTVTGVSASCSHNGP19 145—164 VTGVSASCSHNGKSSFYRNLP20 153—172 SHNGKSSFYRNLLWLTGKNG

HA pool #5 P21 161—180 YRNLLWLTGKNGLYPNLSKS DR0103P22 169—188 GKNGLYPNLSKSYVNNKEKEP23 177—196 LSKSYVNNKEKEVLVLWGVHP24 185—204 KEKEVLVLWGVHHPPNIGNQP25 193—212 WGVHHPPNIGNQRALYHTEN

HA pool #6 P26 201—220 IGNQRALYHTENAYVSVVSS DR1501P27 209—228 HTENAYVSVVSSHYSRRFTPP28 217—236 VVSSHYSRRFTPEIAKRPKV DR0103P29 225—244 RFTPEIAKRPKVRDQEGRINP30 233—252 RPKVRDQEGRINYYWTLLEP

HA pool #7 P31 241—260 GRINYYWTLLEPGDTIIFEAP32 249—268 LLEPGDTIIFEANGNLIAPWP33 257—276 IFEANGNLIAPWYAFALSRG DR1501P34 265—284 IAPWYAFALSRGFGSGIITSP35 273—292 LSRGFGSGIITSNAPMDECD

HA pool #8 P36 281—300 IITSNAPMDECDAKCQTPQGP37 289—308 DECDAKCQTPQGAINSSLPFP38 297—316 TPQGAINSSLPFQNVHPVTIP39 305—324 SLPFQNVHPVTIGECPKYVRP40 313—332 PVTIGECPKYVRSAKLRMVTP41 321—338 KYVRSAKLRMVTGLRNIH

a Bold peptide names and peptide sequences represent T cell epitopes identified.

Mapping of CD4 T cell epitopes 23

concentration of 10 mg/ml. The peptides were distrib-uted into 8 pools of five peptides each (plus p41 as anindividual peptide) for HA and 6 pools of five peptideseach for MP. Other studies have indicated that a poolsize of five peptides is convenient and preserves thesensitivity needed to identify individual peptide epitopes[3,12,17].

HLA class II tetramers

The vector construction, expression, purification, andassembly of class II tetramers were accomplished aspreviously described [11]. The following class II tetramerswere used in this study: DRA*0101/DRB1*0103 (DR0103),

DRA*0101/DRB1*1501 (DR1501), DRA*0101/DRB5*0101(DRB5), and DPA1*0103/DPB1*0401 (DP0401).

Tetramers for screening peptide pools and mappingindividual epitopes were generated as previously described[12].

T cells stimulation and tetramer staining

PBMC from immunized healthy donors were seeded in 24-well plates at 5 � 106cells/well in 2-ml T cell medium(RPMI 1640 with 10% pooled human serum and 100 Ag/mlof penicillin and 100 U/ml streptomycin). The cells werestimulated with 10 Ag/ml of peptide (pooled or individualpeptide as indicated in the text). After culturing for 7

Table 3 Overlapping peptides spanning the influenza A matrix protein (MP)

Peptide pools Peptide name Residues Peptide sequence MHC restriction

MP Pool #1 p42 1—20 MSLLTEVETYVLSIIPSGPLp43a 9—28 TYVLSIIPSGPLKAEIAQRL DR1501

p44 17—36 SGPLKAEIAQRLEDVFAGKNp45 25—44 AQRLEDVFAGKNTDLEVLMEp46 33—52 AGKNTDLEVLMEWLKTRPIL

MP Pool #2 p47 41—60 VLMEWLKTRPILSPLTKGIL DR0103, DP0401

p48 49—68 RPILSPLTKGILGFVFTLTVp49 57—76 KGILGFVFTLTVPSERGLQR

p50 65—84 TLTVPSERGLQRRRFVQNALp51 73—92 GLQRRRFVQNALNGNGDPNN

MP Pool #3 p52 81—100 QNALNGNGDPNNMDKAVKLYp53 89—108 DPNNMDKAVKLYRKLKREITp54 97—116 VKLYRKLKREITFHGAKEIS DR0103

p55 105—124 REITFHGAKEISLSYSAGAL DR0103

p56 113—132 KEISLSYSAGALASCMGLIYMP Pool #4 p57 121—140 AGALASCMGLIYNRMGAVTT

p58 129—148 GLIYNRMGAVTTEVAFGLVCp59 137—156 AVTTEVAFGLVCATCEQIADp60 145—164 GLVCATCEQIADSQHRSHRQp61 153—172 QIADSQHRSHRQMVTTTNPL

MP Pool #5 p62 161—180 SHRQMVTTTNPLIRHENRMV DR0103

p63 169—188 TNPLIRHENRMVLASTTAKA DR0103, DR1501, DRB5

p64 177—196 NRMVLASTTAKAMEQMAGSS DRB5

p65 185—204 TAKAMEQMAGSSEQAAEAMEp66 193—212 AGSSEQAAEAMEVASQARQM

MP Pool #6 p67 201—220 EAMEVASQARQMVQAMRTIGp68 209—228 ARQMVQAMRTIGTHPSSSAGp69 217—236 RTIGTHPSSSAGLKNDLLENp70 225—244 SSAGLKNDLLENLQAYQKRM DRB5

p71 233—252 LLENLQAYQKRMGVQMQRFK DR0103a Bold peptide names and peptide sequences represent T cell epitopes identified.

J. Yang et al.24

days at 378C and 5% CO2, the cells were fed by doublingthe culture volume with fresh medium containing IL-2 (10U/ml final concentration) and split into two wells.Subsequently, the cells were fed with fresh T cellmedium containing IL-2 every 3 days until the end ofthe experiment. Tetramer staining was carried outbetween days 14 and 30 of the culture. For tetramerstaining, 0.2 million cells in 100 Al was stained with 2Al of the appropriate tetramers for 2 h at 378C, 5% CO2.In most experiments, the cells were further stained with5 Al of anti-CD4-FITC at 48C for 10 min. The cells werewashed twice in PBS and analyzed using a FACS Calibur(BD biosciences, San Jose, CA). After the initial round oftetramer screening (screening peptide pools), cells fromtetramer-positive wells were subjected to a secondscreening (fine typing) using sets of five tetramers, eachloaded with one individual peptide from within thecorresponding peptide pool.

Magnetic bead isolation of CD4 cells

In experiments designed to simultaneously screen forepitopes presented on 4 HLA class II alleles simultaneously,CD4 cells were isolated using a no-touch CD4+ purificationkit (Miltenyi Biotech, Auburn, CA) according to the manu-

facturer’s recommendations. Purified CD4+ T cells weresubsequently stained with tetramers labeled with fourdifferent chromophores — each corresponding to a differentclass II allele. Tetramer staining was accomplished asdescribed above, except that the CD4-FITC antibody wasnot used. This approach allowed the use of a fourth channelfor tetramer analysis.

ELISpot assay

For each ELISpot assay, one 96-well PVDF membraneplate (Millipore, Bedford, MA) was coated with anti-Interferon g (Endogen/Pierce, Rockford, IL) at a concen-tration of 0.5 Ag/ml (100 Al per well). PBMC fromhealthy donors within the set used for epitope mappingwere seeded in the ELISpot plate at 4 � 105 cells/welland stimulated with antigenic peptides or non-antigeniccontrol peptides (as determined by TGEM experiments)at 10 Ag/ml. After 48 h of stimulation, the cells wereremoved by washing with PBS and PBS-Tween. IFN-gspots were detected using biotinylated anti-IFN-g (Endo-gen/Pierce, Rockford, IL) followed by avidin—horse radishperoxidase (BD PharMingen, San Diego, CA) and devel-oped with Vectastain AEC substrate (Vector Labs,Burlingame, CA).

Mapping of CD4 T cell epitopes 25

Results

Defining of a set of chromophores for multiplexdetection of T cells with tetramers

To date, MHC class II tetramers have been applied almostexclusively using PE as the fluorescent label. For thisstudy, it was necessary to define a set of chromophoressuitable for simultaneous use in tetramer staining on a 4-color flow cytometry instrument. To this end, we assem-bled tetramers using DR0103/HA306—318 monomer andstreptavidin conjugated with a panel of readily availablechromophores: FITC, PE, PerCP, PerCP-Cy5.5 and APC.PBMC from a healthy donor carrying the HLA-DRB1*0103allele were stimulated with Influenza A virus hemaggluti-nin (HA) peptide HA306—318 for 16 days and subsequentlystained with tetramers labeled with each of the chromo-phores within the panel. As shown in Fig. 1, the hierarchyof fluorescent intensity of tetramer staining is PE N

allophycocyanin (APC) N Peridinin—chlorophyll—proteincomplex—cyanine 5.5 (PerCP-Cy5.5) N Peridinin—chloro-phyll—protein complex (PerCP) N FITC. Because the FITC-labeled tetramer failed to detect HA specific T cells,additional experiments were carried out to identify a labelthat could be used in the FITC channel. Streptavidinlabeled with alexa488 was found to be a better optionthan FITC (data not shown). In subsequent studiesdiscussed below, PE, PerCP, and APC were selected for3-color multiplex studies, and PE, APC, PerCP-Cy5.5 andAlexa 488 were selected as labels for 4-color multiplextetramer analysis.

Simultaneous identification of T cell epitopes onmultiple HLA class II alleles

Because most individuals carry multiple alleles for HLA-DR, DQ and DP, a large amount of epitope informationcould be gleaned from a single experiment provided thatmultiple alleles could be tested simultaneously. Toinvestigate this possibility, we assembled three sets oftetramers corresponding to three of the class II allelesDR0103, DR1501, and DRB5 carried by one of our healthy,flu-immunized donors. Tetramers were generated (asdescribed in Materials and methods) by loading class IImonomers with HA peptide pools (Table 2) and then

Figure 1 Tetramer staining of HA306—318 stimulated PBMC usinghealthy DR0103 donor were stimulated with 10 Ag/ml of HA306—31HA306—318 tetramers labeled with PE, APC, PerCP-Cy5.5, PerCP, or

conjugating with labeled streptavidin. The DR0103 tetra-mers were assembled using streptavidin PE, the DR1501with streptavidin PerCP, and the DRB5 with streptavidinAPC. PBMC from the healthy immunized donor werestimulated with pooled peptides, one well for each pool(2 Ag/ml for each of the five peptides—10 Ag/ml total).Fourteen days post-stimulation, the cells were interrogat-ed with the 3 sets of pooled tetramers simultaneously.These staining results are shown in Fig. 2A. Positivestaining was obtained for peptide pools 5 and 6 forDR0103, peptide pools 6 and 7 for DR1501, and peptidepools 3 and 4 for DRB5. The staining observed wasspecific as cells which were stimulated with tetanustoxoid peptides did not give positive staining (Fig. 2B).To identify the antigenic peptides within the pools, newtetramers were generated using each individual peptidefrom the positive pools and used to stain the samepeptide stimulated PBMC again. As shown in Fig. 3, theresults indicated the presence of T cell epitopes withinthe following peptides: p21 and p28 restricted to DR0103,p26 and p33 restricted to DR1501, p15 and p16 restrictedto DRB5. The peptide sequences of these epitopes areindicated in Table 2. Since the peptides used for epitopemapping studies overlapped by 12 (out of a total of 20)amino acids, it seemed likely that peptides 15 and 16contain an identical epitope. Indeed, when the PBMCfrom same donor were stimulated with peptide p15 only,both HLA-DRB5/p15 and HLA-DRB5/p16 tetramers candetect tetramer-positive cells and vice versa (data notshown).

In the same manner, we assembled tetramers toidentify antigenic epitopes contained in Flu matrixprotein (MP). Tetramers were generated by loading classII monomers with MP peptide pools (Table 3). Tomaximize the amount of information, we assembled foursets of tetramers corresponding to four of the class IIalleles carried by the same healthy, flu-immunized donor(DR0103, DR1501, DRB5, and DP0401). The DR0103 tetra-mers were assembled using streptavidin PE, the DR1501with streptavidin PerCP-Cy5.5, the DRB5 with streptavidinAPC, and the DP0401 with streptavidin labeled with Alexa488. PBMC from the normal donor were stimulated withpooled peptides, one well for each pool (2 Ag/ml foreach of the five peptides—10 Ag/ml in total). Fourteendays post-stimulation, CD4+ T cells were isolated bynegative selection (using Miltenyi Biotec no touch CD4+

streptavidin labeled with various chromophores. PBMC from a

8 peptide for 16 days and subsequently stained with DR0103/FITC.

Figure 2 Simultaneous mapping of antigenic epitopes from HA stimulated cells. (A) PBMC from a healthy immunized donor carrying DR0103, DR1501, and DRB5 were stimulated withHA pooled peptides. Fourteen days post-stimulation, the cells were interrogated with PE-labeled DR0103 (1st row), PerCP-labeled DR1501 (2nd row), and APC-labeled DRB5 (3rd row)tetramer mixtures. (B) Cells from the same subject were stimulated with tetanus toxoid peptides pool #5 [14], and the cells were interrogated with PE-labeled DR0103/HA pool #1 toDR/0103/HA pool #8 tetramers. The percentages of tetramer-positive cells are indicated in the upper right quadrant. Background staining for all experiments was less than 0.2%.

J.Ya

ngetal.

26

Figure 3 Fine mapping of HA epitopes for DR0103, DR1501, and DRB5. The tetramer-positive wells shown in Fig. 2 were stainedwith new tetramers generated using each individual peptide from the six positive pools approximately 1 week after the originalstaining. The percentages of tetramer-positive cells are indicated. Background staining was less than 0.2% for DR0103 and DRB5staining, and less than 0.4% for DR1501 staining.

Mapping of CD4 T cell epitopes 27

T cell isolation kit II) and then interrogated with the 4sets of pooled tetramers simultaneously. These stainingresults are shown in Fig. 4. Positive staining was obtainedfor peptide pools 3, 5, and 6 for DR0103, peptide pools 1and 5 for DR1501, peptide pools 5 and 6 for DRB5, andpeptide pool 2 for DP0401. To identify the antigenicpeptides within the pools, new tetramers were generatedusing each individual peptide from the positive pools andused to stain the corresponding stimulated PMBC again. Asshown in Fig. 5A, the results indicated the presence of Tcell epitopes within the following peptides: p47, p54, p55,p62, p63, and p71 restricted to DR0103, p43 and p63restricted to DR1501, p63, p64, and p70 restricted toDRB5, and p47 restricted to DP0401. The peptide sequen-ces of these epitopes are indicated in Table 3. In a fewoccasions, we observed that a certain proportion oftetramer-positive T cells were CD4 negative (p54, p55and p71 with DR0103). We speculated that this might be aresult of CD4 downregulation after TcR engagement, asall tetramer-positive T cells were CD4 positive if stainingwas carried out for 15 min (Fig. 5B) rather than 120 min(Fig. 5A). As with the epitope found with HA peptidesp15 and p16 for DRB5/HA, it is highly probable that pairs

of consecutive MP peptides — peptides 54 and 55 (restrictedto DR0103), peptides 62 and 63 (restricted to DR0103), andpeptides 63 and 64 (restricted to DRB5) — contain identicalepitopes.

Confirmation of TGEM epitopes by ELISpot assay

To demonstrate that the epitopes identified by multiplexTGEM are reliable, we verified the antigenicity of theMP epitopes identified by TGEM using an interferon g

ELISpot assay. For each ELISpot assay, PBMC from ahealthy immunized donor (also used for MP epitopemapping) were seeded in an ELISpot plate and stimu-lated with antigenic peptides or non-antigenic controlpeptides for 48 h and subsequently developed and readas described in Materials and methods. RepresentativeELISpot results are summarized in Table 4. Sevenantigenic peptides (p47, p54, p55, p62, p63, p70, andp71) and two non-antigenic controls (p1 and p57) weretested for this donor. Consistent with our expectations,the antigenic peptides elicited varying degrees of g-interferon production while the non-antigenic controlselicited almost none. In general, the magnitude of the

Figure 4 Simultaneous mapping of antigenic epitopes from MP stimulated cells. PBMC from the same normal immunized donor(carrying DR0103, DR1501, DRB5, and DP0401) were stimulated with MP pooled peptides. Fourteen days post-stimulation, CD4+ Tcells were isolated by negative magnetic selection and interrogated with Alexa488-labeled DP0401 (top panel, x-axis), PE-labeledDR0103 (top panel, y-axis), PerCP-Cy5.5-labeled DR1501 (bottom panel, x-axis), and APC-labeled DRB5 (bottom panel, y-axis)tetramer mixtures. Background staining was less than 0.2%.

J. Yang et al.28

cytokine response correlated with the observed tetramerresponse.

Multiple antigenic epitopes can be mappedsimultaneously in a single well

In the TGEM system, cells are stimulated using 5 or even10 peptides per well. One potential limitation for thisapproach is that in cases where multiple antigenicepitopes are presented within the same well, T cellexpansion in response to one peptide could interfere withT cell expansion in response to others. For example,stronger antigenic epitopes may stimulate T cell expansionmore efficiently, thereby inhibiting the expansion of Tcells restricted to weaker antigenic epitopes. To investi-gate this scenario, T cells were stimulated with peptidesrestricted to four different class II alleles HA p16 (DRB5),HA p21 (DR0103), HA p33 (DR1501), and MP p47 (DP0401)either together (in one well) or individually (in separatewells). As shown in Fig. 6, panels A (one well) and B(separate wells), multiple antigenic epitopes could still beidentified by tetramers. In fact, when the four antigenicpeptides were presented in one well, the T cells expandedbetter and were detected more easily with tetramers.Alternately in cases where multiple epitopes are pre-sented on the same MHC class II allele, competition foravailable MHC binding sites by the peptides within themixture could favor the presentation of those having ahigh binding affinity—regardless of their antigenicity. Inthis way, one peptide with a high binding affinity couldout-compete or antagonize other peptides with lowerbinding affinities [18]. To investigate the second scenario,T cells were stimulated with four peptides restricted tothe same class II allele: MP p47, p55, p63 and p71 — allrestricted by DR0103 allele — either together (one well) orindividually (separate wells). These results are shown in

Fig. 6 panels C (one well) and D (separate wells). Again,the T cells expanded better and were detected moreeasily when the peptides were presented in a single well—even for p47, which was the weakest epitope. Theseexperiments confirm that multiple epitopes present to-gether in the same well will not impair the identificationof each individual epitope.

It is also possible that the presence of two or more strongbinding peptides within the same peptide pool couldadversely influence the valency of the pooled tetramerreagent. However, previous work in our laboratory demon-strated that the presence of two high affinity MHC bindingpeptides in the same peptide pool did not interfere with theformation of multimers specific for either peptide [19].Though we cannot completely rule out the possibility thatcompetition for binding of peptides to MHC in either antigenpresentation or multimer formation may result in a falsenegative under certain circumstances, the current datadefinitely indicated that multiple antigenic epitopes can bemapped simultaneously in a single well.

Discussions

In our previous work, we developed a tetramer-guidedepitope mapping (TGEM) method to identify the class IIrestricted epitopes of a single antigen for individual class IIalleles. The basis for this TGEM approach is the ability oftetramers to reveal precise class II restricted T cellepitopes directly, using a two-step staining procedure[12,13]. In this way, T cell epitopes could be identified ina single experimental setting within 3 weeks. As eachindividual possesses multiple alleles for HLA-DR, DQ and DP,it would be advantageous to identify antigenic epitopes inthe most expedient and cost-effective manner possible. Wereasoned that a multiplex TGEM assay would be a useful

Figure 5 Fine mapping of MP epitopes for DP0401, DR0103, DR1501, and DRB5. (A) The tetramer-positive wells shown in Fig. 4 werestained with new tetramers generated using each individual peptide from the nine positive pools approximately 1 week after theoriginal staining. Only positive results are shown. (B) Cells as obtained from MP pool 3 were stained with DR0103/MPp54 and DR0103/MPp55 tetramers for 15 min at 378C compared to the 120 min staining protocol as used in panel A. In addition, cells that were beingstimulated with tetanus toxoid peptides did not give positive staining signal with DR0103/MPp54 or DR0103/MPp55 tetramers.

Mapping of CD4 T cell epitopes 29

tool to identify epitopes restricted on multiple (ideally, thecomplete set) HLA class II alleles. Obviously, a moresubstantial amount of epitope information could begleaned from each experiment if multiple alleles couldbe tested using a single sample. The goal of this presentstudy, then, was to demonstrate that the TGEM approachcould be expanded to identify T cell epitopes restricted bymultiple HLA class II alleles simultaneously. Our resultsdemonstrated that multiplex TGEM could be used toidentify T cell epitopes for as many as four class II allelessimultaneously using a distinct fluorescent marker to label

Table 4 Interferon g ELISpot assay data for MP peptides

p1 p57 p47 p54

Restriction Irrelevant Irrelevant DR0103 DP0401 DR0103

Spotsa

(average)1 0 6 42

Tetramer negative negative weak stronga Average from 4 wells with 4 � 105 PBMC/well.

tetramers for each allele. With continuing development inmulti-color Flow cytometry instrumentation and theexpanding availability of soluble class II tetramer reagents,high throughput epitope mapping may become a reality inthe near future.

Specificity and sensitivity of TGEM

The readout for the TGEM assay is direct visualization oftetramer-positive cells by flow cytometry. Thus, the TGEMapproach is highly specific. This differs from other

p55 p62 p63 p70 p71

DR0103 DR0103 DR0103DRB5 DR1501

DRB5 DR0103

42 8 26 23 6

strong moderate strong weak moderate

Figure 6 Simultaneous stimulation and detection of responses to four antigenic peptides restricted to four different class II allelesin one well (panel A) or separate wells (panel B) or restricted to the same class II allele in one well (panel C) or separate wells (panelD). For panels A and B, PBMC from a normal immunized donor carrying the desired class II alleles were stimulated with HA peptides16, 21, and 33 and MP peptide 47-individually or as a mixture. For panels C and D, PBMC from the same donor were stimulated withMP peptides 47, 55, 63, and 71-individually or as a mixture. All wells were stained with the corresponding tetramers to characterizethe response.

J. Yang et al.30

approaches in which the outcome must be inferred fromindirect readouts such as cytokine secretion or proliferation.The TGEM approach is also quite sensitive. Based on CFSEassays (data not shown), we estimate that antigen-specific Tcells can be routinely detected at frequencies as low as 1 in300,000 [20]. Tcells occurring at such low frequencies wouldbe difficult to detect in a standard ELISpot assay. As antigenspecific CD4+ T cells are known to be present at lowfrequencies in the peripheral blood, a sensitive approachsuch as TGEM is essential to reveal the relevant antigenicdeterminants in most settings.

Computer-based MHC epitope prediction and TGEM

In recent years, structural studies of MHC class II-peptidebinding, information about peptide binding preferences,and known class II restricted epitopes have accumulated atan encouraging rate. Based on this emerging information, anumber of computer-based algorithms have been devel-oped to predict potential MHC-binding epitopes withinamino acid sequences [5,6,21—24]. The primary aim ofthese algorithms has been to expedite experimental efforts

to identify the antigenic epitopes for a target antigen bydefining a set of structural motifs that has the highestprobability of binding a given class II allele. However, it isdifficult to assess the sensitivity and selectivity of thesemethods. Raising the selection stringency will reduce thenumber of candidate epitopes, but this will increase therisk of omitting truly antigenic epitopes; lowering theselection stringency will decrease the risk of omitting realepitopes, but this will increase the number of candidateepitopes. Using ProPred (a free epitope prediction programavailable online at http://www.imtech.res.in/raghava/propred/) set at the default threshold of 3% to predictDR1501 restricted epitopes for the HA protein resulted inonly 6 candidate epitopes. However, all of these epitopeswere negative in our tetramer assay. In fact, the motifIAPWYAFAL within the peptide HA p33 that gave thestrongest tetramer response was ranked 11th and themotif LYHTENAYV within the second peptide HA p26 thatgave a tetramer-positive response was ranked 22nd in thepeptide binding scan. The tetramer-positive motifLSSVSSFER for DRB5 ranked 5th (HA p15) and was withinthe default threshold. Predictions were not available for

Mapping of CD4 T cell epitopes 31

DR0103 or DP0401. In contrast to these mixed results, theTGEM method can reveal HA epitopes for a large number ofclass II alleles with the same set of 41 overlappingpeptides, leaving significantly less uncertainty aboutmissed epitopes and false positives.

Implications for MHC binding and TCR engagementfor peptide mixtures

During our initial development of the TGEM strategy, weobserved that tetramers loaded with pooled peptide mix-tures worked effectively to stain the stimulated cells. It hasbeen demonstrated that class II MHC monomers can interactonly weakly with the TCR, while multimers can form asustained interaction [25]. In light of this observation, weexpected that a tetramer must have at least two adjacentclass II molecules bound to the same antigenic peptide tolabel a responding cell. Assuming roughly equal bindingaffinities within a pool of 5 peptides, this would be a rareoccurrence. Our results suggest that true epitopes have ahigher binding affinity when compared to non-antigenicpeptides within the same pool, thereby increasing thelikelihood of adjacent class II molecules binding to the samepeptide. Alternately, it may be that only minimal amounts ofpeptide loaded class II are needed to label the responding Tcells.

Summary

In summary, we have demonstrated that multiplex TGEMprovides a high throughput approach to identify T cellepitopes restricted by multiple class II alleles. Thesefindings significantly expand the utility of the technique,allowing epitope mapping with as many as four distinct setsof tetramers at one time. The TGEM approach provides someadvantages over other approaches — most notably the rapidand robust nature of the assay and the precise determina-tion of the class II restriction element for each epitopeidentified. As class II tetramers become more readilyavailable to the scientific community and as reagents aredeveloped for more class II alleles, the multiplex TGEMapproach should generate significant new information aboutT cell epitopes within foreign proteins. In addition, thetechniques described in this paper may be applicable to themapping of epitopes within autoantigens including tumorantigens.

Acknowledgments

We thank Matt Warren for the secretarial assistance. This workwas supported in part by NIH contract HHSN266200400028Cand the Immune Tolerance Network.

References

[1] D.D. Anthony, P.V. Lehmann, T-cell epitope mapping using theELISPOT approach, Methods 29 (2003) 260–269.

[2] T.W. Tobery, S. Wang, X.M. Wang, M.P. Neeper, K.U. Jansen,W.L. McClements, M.J. Caulfield, A simple and efficientmethod for the monitoring of antigen-specific T cell responses

using peptide pool arrays in a modified ELISpot assay,J. Immunol. Methods 254 (2001) 59–66.

[3] F. Kern, I.P. Surel, C. Brock, B. Freistedt, H. Radtke, A.Scheffold, R. Blasczyk, P. Reinke, J. Schneider-Mergener, A.Radbruch, P. Walden, H.D. Volk, T-cell epitope mapping by flowcytometry, Nat. Med. 4 (1998) 975–978.

[4] B. Hoffmeister, F. Kiecker, L. Tesfa, H.D. Volk, L.J. Picker, F.Kern, Mapping T cell epitopes by flow cytometry, Methods 29(2003) 270–281.

[5] V. Brusic, G. Rudy, G. Honeyman, J. Hammer, L. Harrison,Prediction of MHC class II-binding peptides using an evolution-ary algorithm and artificial neural network, Bioinformatics 14(1998) 121–130.

[6] H. Bian, J.F. Reidhaar-Olson, J. Hammer, The use of bioinfor-matics for identifying class II-restricted T-cell epitopes,Methods 29 (2003) 299–309.

[7] C. Lemmel, S. Stevanovic, The use of HPLC-MS in T-cell epitopeidentification, Methods 29 (2003) 248–259.

[8] M. Sospedra, C. Pinilla, R. Martin, Use of combinatorialpeptide libraries for T-cell epitope mapping, Methods 29(2003) 236–247.

[9] V. Judkowski, C. Pinilla, K. Schroder, L. Tucker, N. Sarvetnick,D.B. Wilson, Identification of MHC class II-restricted peptideligands, including a glutamic acid decarboxylase 65 sequence,that stimulate diabetogenic T cells from transgenic BDC2.5nonobese diabetic mice, J. Immunol. 166 (2001) 908–917.

[10] R.S. Abraham, C.S. David, Identification of HLA-class-II-restricted epitopes of autoantigens in transgenic mice, Curr.Opin. Immunol. 12 (2000) 122–129.

[11] E.J. Novak, A.W. Liu, G.T. Nepom, W.W. Kwok, MHC class IItetramers identify peptide-specific human CD4(+) T cellsproliferating in response to influenza A antigen, J. Clin. Invest.104 (1999) R63–R67.

[12] E.J. Novak, A.W. Liu, J.A. Gebe, B. Falk, G.T. Nepom, D.M.Koelle, W.W. Kwok, Tetramer guided epitope mapping: rapididentification and characterization of immunodominant CD4+ Tcell epitopes from complex antigens, J. Immunol. 166 (2001)6665–6670.

[13] W.W. Kwok, J. Gebe, A. Liu, S. Agar, N. Ptacek, J. Hammer, D.Koelle, G.T. Nepom, Rapid epitope identification from complexclass II restricted T cell antigens, Trends Immunol. 22 (2001)583–588.

[14] J. Yang, L. Huston, D. Berger, N.A. Danke, A.W. Liu, M.L. Disis,W.W. Kwok, Expression of HLA-DP0401 molecules for identifi-cation of DP0401 restricted antigen specific T cells, J. Clin.Immunol. 25 (2005) 428–436.

[15] J. Yang, N.A. Danke, D. Berger, S. Reichstetter, H. Reijonen,C. Greenbaum, C. Pihoker, E. James, W.W. Kwok, Islet-specific Glucose-6 phosphatase catalytic subunit-related pro-tein reactive CD4+ T cells in human subjects, J. Immunol.(2006) 2781–2789.

[16] E. Mickelson, A. Smith, S. McKinney, G. Anderson, J.A. Hansen,A comparative study of HLA-DRB1 typing by standard serologyand hybridization of non-radioactive sequence-specific oligo-nucleotide probes to PCR-amplified DNA, Tissue Antigens 41(1993) 86–93.

[17] F. Kern, E. Khatamzas, I. Surel, C. Frommel, P. Reinke, S.L.Waldrop, L.J. Picker, H.D. Volk, Distribution of human CMV-specific memory T cells among the CD8pos. Subsets defined byCD57, CD27, and CD45 isoforms, Eur. J. Immunol. 29 (1999)2908–2915.

[18] J.D. Hayball, B.W. Robinson, R.A. Lake, CD4+ T cells cross-compete for MHC class II-restricted peptide antigen complexeson the surface of antigen presenting cells, Immunol. Cell Biol.82 (2004) 103–111.

[19] W.W. Kwok, N.A. Ptacek, A.W. Liu, J.H. Buckner, Use of class IItetramers for identification of CD4+ T cells, J. Immunol.Methods 268 (2002) 71–81.

J. Yang et al.32

[20] N.A. Danke, W.W. Kwok, HLA class II-restricted CD4+ T cellresponses directed against influenza viral antigens postin-fluenza vaccination, J. Immunol. 171 (2003) 3163–3169.

[21] J. Hammer, P. Valsasnini, K. Tolba, D. Bolin, J. Higelin, B.Takacs, F. Sinigaglia, Promiscuous and allele-specific anchors inHLA-DR-binding peptides, Cell 74 (1993) 197–203.

[22] J. Hammer, T. Sturniolo, F. Sinigaglia, HLA class II peptide bindingspecificity and autoimmunity, Adv. Immunol. 66 (1997) 67–100.

[23] J. Hammer, E. Bono, F. Gallazzi, C. Belunis, Z. Nagy, F.Sinigaglia, Precise prediction of major histocompatibilitycomplex class II-peptide interaction based on peptide sidechain scanning, J. Exp. Med. 180 (1994) 2353–2358.

[24] T. Sturniolo, E. Bono, J. Ding, L. Raddrizzani, O. Tuereci,U. Sahin, M. Braxenthaler, F. Gallazzi, M.P. Protti, F.Sinigaglia, J. Hammer, Generation of tissue-specific andpromiscuous HLA ligand databases using DNA microarraysand virtual HLA class II matrices, Nat. Biotechnol. 17(1999) 555–561.

[25] J.J. Boniface, J.D. Rabinowitz, C. Wulfing, J. Hampl, Z.Reich, J.D. Altman, R.M. Kantor, C. Beeson, H.M. McCon-nell, M.M. Davis, Initiation of signal transduction throughthe T cell receptor requires the multivalent engagementof peptide/MHC ligands, (corrected)Immunity 9 (1998)459–466.