CTL recognition of a novel HLA-A*0201-binding peptide derived from glioblastoma multiforme tumor cells

ORIGINAL ARTICLE

CTL recognition of a novel HLA-A*0201-binding peptide derivedfrom glioblastoma multiforme tumor cells

Cheryl E. Myers • Paul Hanavan • Kwasi Antwi • Daruka Mahadevan •

A. Jamal Nadeem • Laurence Cooke • Adrienne C. Scheck • Zachary Laughrey •

Douglas F. Lake

Received: 24 March 2010 / Accepted: 9 May 2011 / Published online: 29 May 2011

� Springer-Verlag 2011

Abstract Genetic instability of tumor cells can result in

translation of proteins that are out of frame, resulting in

expression of neopeptides. These neopeptides are not self-

proteins and therefore should be immunogenic. By eluting

peptides from human glioblastoma multiforme (GBM)

tumor cell surfaces and subjecting them to tandem mass

spectrometry, we identified a novel peptide (KLWGL

TPKVTPS) corresponding to a frameshift in the 30 beta-

hydroxysteroid dehydrogenase type 7 (HSD3B7) gene.

HLA-binding algorithms predicted that a 9-amino acid

sequence embedded in this peptide would bind to HLA-

A*0201. We confirmed this prediction using an HLA-

A*0201 refolding assay followed by live cell relative

affinity assays, but also showed that the 12-mer binds to

HLA-A*0201. Based on the 9-mer sequence, optimized

peptide ligands (OPL) were designed and tested for their

affinities to HLA-A*0201 and their abilities to elicit anti-

peptide and CTL capable of killing GBM in vitro. Wild-

type peptides as well as OPL induced anti-peptide CTL as

measured by IFN-c ELISPOTS. These CTL also killed

GBM tumor cells in chromium-51 release assays. This

study reports a new CTL target in GBM and further sub-

stantiates the concept that rational design and testing of

multiple peptides for the same T-cell epitope elicits a

broader response among different individuals than single

peptide immunization.

Keywords Cytotoxic T lymphocytes � Optimized peptide

ligand � Glioblastoma multiforme

Introduction

Nearly 11,000 individuals were diagnosed with glioblas-

toma multiforme (GBM) in 2009. Statistics suggest that

less than 30% of those patients will be alive 5 years after

diagnosis. Treatment for GBM includes neurosurgery,

chemotherapy, and radiotherapy, but the tumor almost

always recurs, likely due to chemo and radio-resistant stem

cells and/or the inability to resect the entire tumor due to its

location [1, 2]. Obviously, additional studies and approa-

ches directed toward cure and therapy are needed for

patients with GBM.

Although many tumor antigens have been discovered

and characterized [3], there remains a need to identify more

tumor antigens due to the genetic instability and hetero-

geneity of tumors. Over-expressed, tissue-specific, and

mutated proteins each have potential to harbor peptide

epitopes that could stimulate anti-tumor CTL. Character-

ization of peptides that bind to major histocompatibility

complex (MHC) class I has revealed that most peptides are

presented to CD8 T cells as 8–10 amino acid peptides.

However, identification of longer class I T-cell epitopes has

been reported [4–6]. HLA-A*0201 (HLA-A2)-binding

peptides are typically 8–10 amino acids in length with

C. E. Myers � P. Hanavan � K. Antwi � D. F. Lake (&)

School of Life Sciences, Arizona State University, Tempe, AZ

85287, USA

e-mail: [email protected]

Z. Laughrey

Department of Chemistry and Biochemistry,

Arizona State University, Tempe, AZ 85287, USA

D. Mahadevan � A. J. Nadeem � L. Cooke

Department of Medicine, Arizona Cancer Center,

University of Arizona, Tucson, AZ 85724, USA

A. C. Scheck

Barrow Neurological Institute of SJHMC,

Phoenix, AZ 85013, USA

123

Cancer Immunol Immunother (2011) 60:1319–1332

DOI 10.1007/s00262-011-1032-4

anchor residues at positions P2 and 9 or 10 [7, 8]. Positions

2 and 9 anchor peptides in the HLA-A2-peptide-binding

groove, but auxiliary anchor residues at other positions are

also critical for optimal binding to HLA-A2 [9–14]. Longer

peptides can be accommodated in the MHC class I binding

groove by extending through the carboxy-terminus, bulg-

ing out of the binding cleft or zigzagging across the pep-

tide-binding groove [5, 6]. It is not clear how peptides

longer than 8–10 amino acids are loaded onto MHC class I

molecules. One may speculate that longer peptides loaded

onto class I missed trimming by proteases either before or

after the peptide-MHC-b2 microglobulin (b2 m) complex

has exited from the endoplasmic reticulum. We recently

reported that a large portion of peptides longer than 8-10

amino acids are found on the surfaces of tumor cells,

presumably bound to MHC [4].

Mutations in coding regions may result in translation of

proteins that are out of frame [15–18], but usually truncate

with a stop codon resulting in expression of frameshifted

neopeptides [15–19]. In this study, we identified a frame-

shift (FS) peptide derived from the 3 beta-hydroxysteroid

dehydrogenase type 7 (HSD3B7) gene that binds to the

HLA-A*0201 molecule. HSD3B7 plays a role in the initial

stages of the biosynthesis of bile acids from cholesterol

[20]. It is not known to play any role in tumor development

or suppression. Our results reveal that a new antigenic

12-mer peptide present in brain tumor cell lines is capable

of stimulating CTL. The frameshift and optimized peptide

ligands represent potential tumor-specific antigens that can

be recognized by CTL. Since these frameshift peptides are

expressed in an alternate frame and have nothing to do with

the ‘‘self’’-protein, they should be foreign to the immune

system and elicit an immune response if one were to

immunize with them. Subsequent study of this peptide by

other groups will help to validate its potential for

immunotherapy.

Materials and methods

Acid elution of peptides

1.2 9 108 GBM cells from T98g and CRL2610 were

harvested with Cell Stripper (Mediatech Inc., Manassas,

VA). Cells were centrifuged at 2339g (1,000 rpm) for

5 min and washed in 45 ml phosphate-buffered saline

followed by resuspension in 1 ml cold citric acid buffer,

pH 3.0 [21]. After 5 min at low pH, the acid supernatant

was immediately filtered through a 0.45-lm low protein-

binding filter (Pall Corporation) and then through a 3-kDa

ultrafilter (Millipore, Billerica, MA) at 13000 RPM at 4�C.

The filtered samples were then stored at -20�C in

siliconized tubes until they were analyzed by reverse-phase

high-performance liquid chromatography (RP-HPLC)

coupled to tandem mass spectrometry (MS/MS). LC–MS/

MS was performed as previously described except that the

filtered acid elutions were run on a Thermo-Finnigan LTQ

mass spectrometer [4]. Peptide sequences were identified

by searching spectra against a frameshift database con-

structed by translating the coding sequences from the

NCBI CDS database in ?1 and ?2 frames such that it

contained all possible frameshift peptides longer than 7

amino acids. Spectrum Mill software Rev A.03.03.078

(Agilent Technologies) was used to perform spectra

searches.

Peptides synthesized on TentaGel beads

The following peptides were synthesized on TentaGel S NH2

resin beads (Advanced ChemTech) with a substitution rate of

0.2–0.5 mmol/g and were used in this study: KLWG-12

(KLWGLTPKVTPS) and KLWG-9 (KLWGLTPKV), hep-

atitis B core antigen (FLPSDCFPSV, residues 18–27), and

predicted HLA-B*3901 (NHCQLLKVMV).

HLA-A*0201-binding synthetic peptides

The following HLA-A*0201-binding peptides were used in

this study: KLWG-12 (KLWGLTPKVTPS) and KLWG-9

(KLWGLTPKV); OPL1 (FLWGLTPKV) and OPL2

(FLFWGLTPKV); HIV-1 p17 Gag77-85 epitope (SLYNT-

VATL); and influenza virus matrix peptide (M1: 58–66,

GILGFVFTL) as a positive control for CTL generation

experiments. Peptides were synthesized using standard

FMOC chemistry and purified by HPLC to[90% purity.

Detection of KLWG splice variants from tumor cell

lines and tumor tissue

Total RNA was isolated from 1 9 107 GBM cell lines

CRL-2610, T98g, U-87, and U251 using the RNeasy Mini

Kit (Qiagen, Valencia, CA). An optional on-column DNA

digestion was performed with RNase-Free DNase Set

(Qiagen) for 15 min at room temperature. RNA was eluted

from columns using 50 ll DNase/RNase-free Molecular

Grade Water. RNA from primary brain tumor and normal

brain tissue was graciously provided by Dr. Adrienne

Scheck of Barrow Neurological Institute (BNI), Phoenix,

Arizona.

cDNA synthesis was performed using 1 lg total RNA

that was combined with DEPC-treated water and oligo-dT

primer to a total volume of 25 ll in a 0.5-ml microfuge

tube and incubated at 70�C for 10 min, followed immedi-

ately by 50�C incubation (during which reaction mixtures

1320 Cancer Immunol Immunother (2011) 60:1319–1332

123

were prepared). DEPC-treated water, 10X PCR buffer,

MgCl2, dNTP mix, and Dithiothreitol (DTT) were com-

bined and pre-warmed to 42�C and added to the 50�C RNA

mix. To each cDNA synthesis tube, 200 units of Super-

Script II reverse transcriptase (Invitrogen) was added.

Samples were incubated at 50�C for 50 min, and the

reaction was terminated by 70�C incubation for 15 min.

Samples were immediately chilled followed by addition of

1 ll RNase H and incubation for 20 min at 37�C.

Touchdown was performed as described previously

[22]. Reaction mixes of 50 ll consist of 1X PCR buffer

(Promega), 5% DMSO, 200 lM dNTP mix (Promega), 2

units GoTaq (Promega), 200 nM primers (forward, 50-AG

AAGCTGGTGTACCTGGTC-30; reverse, 50-CTGCTTCG

TATGGGGTGTCT-30) and 40 ng cDNA per reaction; 10

touchdown cycles were performed at -1�C per cycle from

68�C to 58�C, and 25 additional cycles were performed at

58�C annealing temperature and 90 s denaturation/exten-

sion times.

PCR products were electrophoresed on 1% agarose gels,

stained with ethidium bromide, and photographed using a

digital camera.

Cell lines and media

T2 cells (HLA-A*0201?) were used as antigen-presenting

cells (APC) for ELISPOT and chromium-51 release assays.

T2 cells were cultured in RPMI 1640 (Mediatech, Inc.),

supplemented with 10% heat-inactivated fetal bovine

serum (Gibco) with 25,000 U/ml Penicillin–Streptomycin

and 29.2 mg/ml L-glutamine (Lonza, Walkersville, MD).

Glioblastoma cell lines, T98g, CRL2610, and U-87

(molecularly typed, HLA-A*0201) were cultured in

DMEM supplemented with 10% fetal bovine serum with

25,000 U/ml Penicillin–Streptomycin and 29.2 mg/ml

L-glutamine (Lonza, Walkersville, MD) and 250 lg/ml

Amphotericin B (MP Biomedicals, Aurora, CA). Cells

were incubated at 37�C and 5% CO2.

Reconstitution of HLA-A*0201 for peptide screening

Refolding of MHC onto peptide beads was performed as

published previously [14, 23]. Briefly, KLWG-12 and

KLWG-9 peptides were synthesized on resin support

beads. Baculovirus-produced recombinant HLA-A*0201

(kind gift from Dr. Margaret Smith) was denatured in

6 M urea and added to approximately 200 peptide beads

synthesized on resin beads. To allow the peptide-MHC-

b2m complex to assemble on the peptide beads, 100 lg of

purified recombinant b2-microglobulin (kind gift from

Dr. Margaret Smith) was added. The mixture was then put

into a Tube-O-Dialyzer (G-Biosciences, St Louis, MO)

with a 1-kDa MWCO pore size. This was dialyzed against

5 M urea in PBS with a 1 M step-down of the urea

concentration every hour until the peptide-bead-HLA-b2m

refolding buffer was exchanged to PBS containing 0.1%

Tween 20 (PBST). After several washes of PBST, the

resin was then incubated with 1 lg/ml anti-class I mAb

W6/32 for 1 h at room temperature, followed by 2 more

washes with PBST. The peptide beads were then incu-

bated in 1:5000 goat anti-mouse IgG conjugated to

alkaline phosphatase (Jackson Immuno Research) for 1 h

at room temperature. The resin was washed 5 times with

PBST. After a final wash of tris-buffered saline (TBS),

100 uM 5-bromo-4chloro-3-indolyl phosphate (Sigma)

was added to the beads as a colorometric enzyme sub-

strate [24, 25]. Peptide beads that turned blue had HLA-

A*0201 refolded onto them. FLPSDCFPSV peptide is

altered HLA-A*0201 epitope from hepatitis B virus

(HBV) core antigen residues 18–27. This HBV peptide

has been used successfully as a positive control for HLA-

A*0201 binding previously [4, 26, 27]. NHCQLLKVMV

peptide is a predicted HLA-B*3901 binder and has been

used previously as a negative control for HLA-A*0201

binding [4].

HLA-A*0201 binding affinity of frameshift peptides

and optimized peptide ligands

Relative affinity assays were performed to determine

affinities of peptides for HLA-A*0201 molecules as pre-

viously reported [26]. The HBV core antigen

(FLPSDYFPSV) was synthesized with a cysteine residue

substituted for the tyrosine residue at position 6. The cys-

teine residue was conjugated to fluorescein (Fl-peptide) for

flow cytometric detection (FLPSDCFPSV: Fl-peptide). T2

cells (5 9 105 cells/tube) were cultured in serum-free

RPMI with recombinant b2-m (2 lg/ml) and Fl-peptide

(0.1 lg/ml), and incubated for 18–20 h with varying con-

centrations of OPL peptides at 26�C in a 5% CO2 incu-

bator. Mean fluorescence intensity (MFI) values were used

to determine inhibition of the Fl-peptide from binding to

HLA-A*0201 molecules on T cells by optimized peptide

ligands. Percent inhibition was calculated as: [1 - (MFI

T2 ? Fl-peptide ? modified peptide - MFI T2 only)/

((MFI T2 ? Fl-peptide) - (MFI T2 only))] 9 100. The

IC50 values of the OPL and KLWG-9 and 12 peptides were

determined by calculating the concentration of peptide

required to inhibit binding of 50% of the Fl-peptide binding

to T2 cells.

Generation of peptide-specific CTL

PBMC were obtained from normal donors by peripheral

blood draws according to the guidelines set forth by the

Human Subjects Committee at the University of Arizona

Cancer Immunol Immunother (2011) 60:1319–1332 1321

123

and Arizona State University. PBMC were purified using

standard white blood cell separation by density centrifu-

gation with Ficoll Hypaque. Peptide-specific CTL were

generated as follows. Dendritic cells (DC) were generated

by plating HLA-A2? PBMC (1 9 107–1 9 108 cells) in a

T-75 flask in 10 ml of AIM-V medium and incubated at

37�C in a 5% incubator for 2 h. Non-adherent cells were

removed and replaced with 10 ml X-Vivo medium. IL-4

(PeproTech Inc, Rocky Hill, NJ) was added at a concen-

tration of 15 ng/ml along with 1,000 IU/ml Leukine

(Berlex, Seattle, WA) to culture medium. On day 5, IL-4,

GM-CSF, TNF-a (500 IU/ml, R&D Systems, Minneapolis,

MN), and prostaglandin E2 (2 lg/ml, Sigma) were added to

culture medium. On day 7, DC were removed from the

flask and resuspended in Iscove’s modified Dulbecco’s

medium (cIMDM) supplemented with 10% heat-inacti-

vated human AB serum (Gemini Bioproducts, Calabasas,

CA), 25,000 U/ml Penicillin–Streptomycin and 29.2 mg/

ml L-glutamine (Lonza, Walkersville, MD) and 250 lg/ml

Amphotericin B (MP Biomedicals, Aurora, CA) in a

24-well plate. Non-adherent autologous lymphocytes were

added to DC cultures (10:1 ratio of effector cells to stim-

ulator cells). To the DC-PBMC cultures, 10 ng/ml of IL-7,

(R&D Systems, Minneapolis, MN) and 100 IU/ml of IL-2

(PeproTech Inc, Rocky Hill, NJ) were added. KLWG and

OPL were added at 1 lg/ml. Every 3–4 days, 100 IU/ml of

IL-2 was added. Cells were split as needed and prolifera-

tive blasts were frequently dissociated. Weekly peptide

stimulations began seven days after the first stimulation.

Autologous PBMC were irradiated (4,000 rads) and pulsed

with 1 lg/ml peptide. At the end of each stimulation, cells

were analyzed by ELISPOT for peptide recognition.

Detection of IFN-c-producing cells by ELISPOT

Seven days after the third peptide stimulation, PBMC

recognition of KLWG peptides was measured by IFN-cELISPOT. T2 cells (2.5 9 104 cells/well) were loaded with

an HLA-A*0201-binding peptide (10.0 lg/ml) and pan

MHC class I blocking antibody, and W6/32 monoclonal

antibody was added as a control for MHC-dependent CTL

activity prior to incubation with effector cells. Effector

cells (5 9 104 cells/well) and peptide-pulsed T2 cells

(2.5 9 104 cells/well) were added to anti-IFN-c antibody-

coated plates in 200 ll IMDM. Control wells contained un-

stimulated PBMC, PBMC ? PHA (10.0 lg/ml) or

PBMC ? T2 cells. After incubation for 36 h at 37�C in 5%

CO2, cells were removed by washing five times with PBS/

0.05% Tween 20. Captured cytokine was detected by

incubation for 5 h at room temperature with biotinylated

mouse anti-human IFN-c monoclonal antibody (2.5 lg/ml/

well, Pharmingen, San Diego, CA) diluted in PBS/0.05%

Tween 20/0.1% FCS. Plates were washed with PBST

followed by addition of streptavidin-HRP (Pharmingen,

San Diego, CA), diluted 1:1000 in PBST-FCS and incu-

bated for 1 h at room temperature. Staining was performed

with 3-amino-9-ethylcarbazole for 15–20 min at room

temperature. Color development was stopped by thor-

oughly rinsing the plates with distilled water. Spots were

counted with an ELISPOT reader using Immunospot soft-

ware (Cellular Technology Ltd, Cleveland, OH).

Cytolytic activity of PBMC

Standard 51Cr release cytotoxicity assays were performed to

evaluate the ability of CTL generated from PBMC to lyse

target cells pulsed with KLWG-12 and KLWG-9 peptide and

brain tumor target cells. T2 cells (5 9 103 cells/well) were

labeled with 100 lCi of 51Cr (Amersham Pharmacia Biotech

Inc., Piscataway, NJ) for 45 min followed by incubation with

10.0 lg/ml of KLWG-12 or KLWG-9 in cIMDM for 1 h.

Prior to assay, W6/32 mAb was added to T2 cells pulsed with

peptide and tumor cell lines as a control for MHC-dependent

CTL activity. HLA-A2? T98g was a kind gift from

Dr. Walter Storkus. CRL2610 and U-87 were purchased

from American Type Culture Collection, Rockville, MD.

K562 cell line served as an NK cell target. Release of

radioactivity was measured by gamma counting (Top Count,

Packard, Meriden, CT). Maximum release was defined by51Cr released from targets lysed with 10% Triton X-100.

Percent specific lysis was calculated as: ((experimental51Cr release - spontaneous release)/(maximum release

- spontaneous release)) 9 100.

Mode of binding of 9-mer and 12-mer peptides

to HLA-A*0201

MODPROPEP evaluation of the 9-mers was based on a

combination of 26 peptide-HLA-A*0201 crystal structures

which indicated that peptide position 2 (L) and position 9

(V) are invariant and interact in a highly homologous

manner with the peptide-binding site [28]. Further,

MODPROPEP provides binding scores based on peptide-

protein interactions specified by closest neighbor (6 A

cutoff) that encompasses hydrogen bonding, hydrophobic,

aromatic, and van der Waals interactions (data not shown).

Moreover, MODPROPEP also provides a pdb file with

peptide threaded into the MHC class I binding site. The

program MODPROPEP is able to thread 8-mers and

10-mers based on crystal structures; however, no structures

exist for peptides [10 residues bound to HLA class I.

Hence, for the 12-mer, the strategy was to learn from

MODPROPEP to interactively dock utilizing Surflex-Dock

(Sybyl V.8, Tripos) with positions 2 and 9 fixed in their

respective hydrophobic pockets and then optimize binding

based on known protein-peptide interactions.


123

Results

Glioblastoma acid eluate contains HLA-A*0201

binding peptides

Peptides derived from MHC class I complexes expressed on

the cell surfaces of T98g and CRL2610 were isolated after

cells were treated with pH 3.0 buffer. Acid eluates from tumor

cells were filtered through 3-kDa MW cutoff spin filters and

then subjected to LC–MS/MS on a Thermo-Finnigan Sur-

veyor LC system (Thermo Electron Corp., San Jose, CA)

online with a linear ion trap (LTQ, Thermo Electron Corp.,

San Jose, CA). Peptide sequences were identified by searching

MS/MS spectra against Swissprot and a custom frameshift

protein database. This frameshift database was constructed by

translating the coding sequences from the NCBI CDS data-

base in ?1 and ?2 frames such that it contained all possible

frameshift peptides longer than 7 amino acids.

A total of 90 peptides were eluted and identified using

LC–MS/MS from T98g; 66 peptide sequences matched

proteins in the Swissprot database and 24 matched our cus-

tom frameshift database. However, only 11 peptides from the

Swissprot database (Table 1A), and 7 peptides from our

custom frameshift database (Table 1B) were predicted to

bind HLA-A*0201 by SYFPEITHI (Table 1A) [29]. Fifty-

five peptides were identified by LC–MS/MS from CRL2610

eluate. Thirty-eight peptides matched proteins in the

Swissprot database and 17 peptides corresponded to our

frameshift database. Eight peptides from Swissprot database

(Table 1C) and 6 peptides from the frameshift database

(Table 1D) were predicted to bind HLA-A*0201 by SYF-

PEITHI [29]. Peptides eluted from the cell lines ranged in

size from 9 to 19 amino acids. Peptides larger than 10 amino

acids were predicted to contain internal HLA-A*0201 epi-

topes as shown in Table 1.

One of the frameshift peptides eluted from T98g,

KLWGLTPKVTPS 12-mer, contained an HLA-A*0201

epitope (KLWGLTPKV) and was predicted to bind to

HLA-A*0201 very strongly with a top score of 30. This

sequence was derived from the ?2 frame of the HSD3B7

protein. Neither KLWG-12 nor KLWG-9 was identified in

acid elutions from CRL2610. Because it was possible that

KLWG-12 was misidentified by the mass spectrometer, we

chemically resynthesized KLWG-12 and analyzed it by

LC–MS/MS, demonstrating the same ion fragmentation in

the mass spectra of the synthetic peptide compared with

that of the natural, cell-derived peptide (Fig. 1).

Detection of HSD3B7 splice variants in clinical

samples

Primers flanking the KLWG peptide region were used to

amplify cDNA synthesized from polyadenylated RNA

from 4 established GBM cell lines, 5 GBM tumor tissues,

one primary low passage GBM line and one normal brain

tissue. In each PCR amplification, a 438-bp product was

observed, suggesting amplification of exons 2, 3, and 4 as

would be expected from NCBI databases. However, in

established and primary tumor cell lines and GBM tumor

tissues, a second band was observed at 282 bp (Fig. 2).

This 282-bp band was not observed in normal brain tissue

(Fig. 2b, lane 7). Sequence analyses indicated that the

438-bp band contains exons 2, 3, and 4 from HSD3B7,

while the 282-bp band is missing exon 3. Cloning of the

282 bp PCR product and further sequence analysis did not

reveal insertions, deletions, or aberrant splicing between

exons 2 and 4 that would have resulted in translation of the

KLWG peptide. None-the-less, all GBM tumor lines and

tissues tested express mRNA containing an HSD3B7 splice

variant missing exon 3, while normal brain appears not to

express this variant.

HLA-A2 Allele Specific Peptide Binding

To validate the SYFPEITHI algorithm-based HLA-A*0201

binding motif prediction for the HSD3B7 frameshift pep-

tide, baculovirus-produced recombinant HLA-A*0201 was

refolded onto ‘‘KLWG-9’’ and ‘‘KLWG-12’’ peptide beads

as previously described [4, 14, 30]. Since

KLWGLTPKVTPS was acid-eluted from cells, we

hypothesized that the 12-mer would bind to HLA-A*0201

based on our previous findings that it is not uncommon for

peptides longer than 8-10 amino acids to bind class I

molecules [4]. Both KLWG-9 and KLWG-12 peptides

were synthesized onto polyethyleneglycol beads (Tentagel)

and incubated with denatured recombinant HLA-A*0201

and b2m, while 6 M urea was dialyzed out of the peptide-

bead/HLA/b2m mixture to re-nature the trimeric complex.

Monoclonal antibody (mAb), W6/32, was then incubated

with peptide beads followed by the addition of alkaline

phosphatase conjugated to goat anti-mouse IgG antibody

and BCIP substrate. Only properly folded MHC trimeric

complexes recognized by mAb W6/32 turn blue. Both

KLWG-12 and KLWG-9 peptide beads demonstrate their

specificity to the HLA-A*0201 molecule as shown by blue

staining of peptide beads (Fig. 3). A control peptide that is

predicted to bind HLA-B*3901 shows no staining.

Relative affinity of HSD3B7 Frameshift peptide

for HLA-A2

Affinity of peptides for HLA molecules is a function of

their immunogenicity. To assess relative affinity of

KLWG-9, KLWG-12, and two optimized peptide ligands,

FLWGLTPKV (OPL1) and FLFGLTPKV (OPL2), IC50

values of the peptides were obtained. Using the T2 cell


123

Table 1 Peptide sequences of HLA-A*0201-binding peptides and epitopes eluted from: a T98g Swissprot database, b T98g frameshift database,

c CRL2610 Swissprot database, d CRL2610 frameshift database

Epitope SYFPEITHIscore Protein name

(A) T98g Parent peptide (length) Swissprot database

VLLESEQFL [9] VLLESEQFL 25 Signal recognition particle 14 kDa protein

DFGVSDADIQEL [12] GVSDADIQEL 22 THO complex subunit 4

RVYLGASTPDLQ [12] YLGASTPDL 25 Sulfiredoxin-1

GSLAIKRDPKVND [13] LAIKRDPKV 21 Low-density lipoprotein receptor-related protein 8

precursor

GSLAIKRDPKVND [13] SLAIKRDPKV 27 Low-density lipoprotein receptor-related protein 8

precursor

DIAVDGEPLGRVSF [13] AVDGEPLGRV 22 Peptidyl-prolyl cis–trans isomerase A

MNLGGLAVARDDGL [14] GLAVARDDGL 24 Transgelin-2

LEGKVLPGVDALSNI [15] KVLPGVDAL 26 Phosphoglycerate kinase 1

EQEMATAASSSSLEKS [16] ATAASSSSL 20 ANKRD26-like family C member 1A

IVNTNVPRASVPDGFLSE [18] NTNVPRASV 20 Macrophage migration inhibitory factor

APVSGPVGLLGARRAWDLE [19] LLGARRAWDL 22 G patch domain-containing protein 8

(B) T98g Parent peptide (length) frameshift database

LLCHQPVASRA [11] LLCHQPVASR 20 Prokineticin-1 precursor

KLWGLTPKVTPS [12] KLWGLTPKV 30 3 beta-hydroxysteroid dehydrogenase type 7

IMKTDKLINLVL [12] KTDKLINLV 21 Zinc finger FYVE domain-containing protein 9

IMKTDKLINLVL [12] IMKTDKLINL 23 Zinc finger FYVE domain-containing protein 9

DLRQEAPADLPGDR [14] LRQEAPADL 22 G-protein-signaling modulator 1

AKLLTIPQTLLNIS [14] KLLTIPQTL 25 Zinc finger protein 493

AKLLTIPQTLLNIS [14] LLTIPQTLL 22 Zinc finger protein 493

AKLLTIPQTLLNIS [14] TIPQTLLNI 21 Zinc finger protein 493

AKLLTIPQTLLNIS [14] KLLTIPQTLL 23 Zinc finger protein 493

AKLLTIPQTLLNIS [14] LTIPQTLLNI 23 Zinc finger protein 493

GRTGGEELRKLLGRE [15] RTGGEELRKL 22 TCF7L2 protein

HGWMASLLRVPGYVLS [16] SLLRVPGYV 24 Protein FAM131B

HGWMASLLRVPGYVLS [16] LLRVPGYVL 24 Protein FAM131B

HGWMASLLRVPGYVLS [16] SLLRVPGYVL 25 Protein FAM131B

(C) CRL2610 Parent peptide (length) Swissprot database

VLLESEQFL [9] VLLESEQFL 25 Signal recognition particle 14 kDa protein (SRP14)

(18-kDa Alu RNA-binding protein)

LLRELKHPNI [10] LLRELKHPNI 23 Serine/threonine-protein kinase Nek2 (EC 2.7.11.1)

(NimA-related protein kinase 2) (NimA-like protein

kinase 1) (HSPK 21)

KVNQIGSVTESI [12] QIGSVTESI 20 Enolase 3 (EC 4.2.1.11) (2-phosphoglycerate

dehydratase 3) (2-phospho-D-glycerate hydro-lyase 3)

DIAVDGEPLGRVS [13] AVDGEPLGRV 22 Peptidyl-prolyl cis–trans isomerase A (PPIase A)

(Rotamase A) (EC 5.2.1.8) (Cyclophilin A)

(Cyclosporin A-binding protein)

EGKVLPGVDALSNI [14] KVLPGVDAL 26 Phosphoglycerate kinase 1 (EC 2.7.2.3)

MNLGGLAVARDDGL [14] GLAVARDDGL 24 Transgelin-2 (SM22-alpha homolog)

MSNLDSNRDNEVDF [14] NLDSNRDNEV 23 Protein S100-A4 (S100 calcium-binding protein A4)

(Metastasin)

LEGKVLPGVDALSNI [15] KVLPGVDAL 26 Phosphoglycerate kinase 1 (EC 2.7.2.3)

(D) CRL2610 Parent peptide (length) frameshift database

TQRLPLLQTL [10] QRLPLLQTL 21 Mucin-19

LLCHQPVASRA [11] LLCHQPVASR 20 Prokineticin-1

LPRGAVPAVRGG [11] LPRGAVPAV 20 Smad nuclear-interacting protein 1

AKLLTIPQTLLNIS [14] KLLTIPQTL 25 Zinc finger protein 493

AKLLTIPQTLLNIS [14] LLTIPQTLL 22 Zinc finger protein 493

AKLLTIPQTLLNIS [14] TIPQTLLNI 21 Zinc finger protein 493


123

affinity assay [27], test peptides compete with a standard

fluoresceinated peptide (Fluor-HBV core Ag18–27) for

binding to HLA-A*0201 on the surfaces of T2 cells. Rel-

ative affinities calculated from the T2 cell assay are shown

in Table 2 along with the half-time dissociation predictions

for each peptide binding to HLA-A*0201 using the SYF-

PEITHI algorithm [29]. Surprisingly, wild-type KLWG-12

peptide (KLWGLTPKVTPS) showed an IC50 of 32 uM for

HLA-A2 on the T2 cell surface, 3-fold better than KLWG-

9 of 110 uM. OPL2 (FLFGLTPKV) demonstrated the best

relative affinity among the 4 peptides tested with an IC50 of

17 uM. Based upon the relative affinity results alone, one

might predict that OPL2 would elicit more active CTL

compared with wild-type KLWG-9. However, we have

observed in previous studies that optimized peptides with

the best affinity do not always elicit the most active CTL

from the most individuals [31]. Therefore, we first modeled

the peptides in HLA-A*0201 and then raised CTL on

KLWG-9, OPL1, and OPL2 peptides followed by testing

their reactivities against wild-type KLWG-9 and KLWG-

12 peptides in cytotoxicity assays.

Modeling of KLWG Peptides in HLA-A*0201

Since KLWG-12 bound to HLA-A*0201 with a better

relative affinity than KLWG-9, we modeled KLWG-9,

KLWG-12, OPL1, and OPL2 in the binding cleft of HLA-

A*0201 in an attempt to observe the potential reason(s) for

the apparent increased affinity of the 12-mer (Fig. 4). The

‘‘Total Score’’ value (Table 2) generated by the modeling

software (higher scores indicate stronger binding) for

KLWG-9 peptide mirrors the IC50 data. Modeling suggests

that KLWG-9 is the poorest binder among the four peptides

essentially due to lack of aromatic residues in P1 like

phenylalanine which are preferred [9, 14]. For KLWG-9,

the amine group in the K at P1 forms a salt bridge with E63

on HLA-A2 and an H-bond with the backbone –C=O of

E58 and Y59, while P3 (W) forms an H-bond with –C=O

backbone of Q155 (Fig. 4a). Residues 5–8 are generally

oriented toward solvent, similar to other HLA-A*0201-

binding peptides.

Although no crystal structures have been generated with

[10 amino acids bound to HLA molecules, our docking

models clearly demonstrate that a 12-mer can bind in the

same orientation as the KLWG-9-mer, and ranks second

best in IC50 measurements and ‘‘Total Score’’ modeling

values among other peptides tested (Table 2). The 12-mer

KLWGLTPKVTPS essentially maintains a similar mode of

binding as KLWG-9 with the N-terminal –NH2 group

forming an H-bind with –OH of Y7 and Y171 on the MHC

molecule. P1 (K) forms an H-bond with –C=O backbone of

Y59 and salt bridge with E63 on the MHC molecule, like

KLWG-9 (Fig. 4b). Unlike KLWG-9, P3 (W) points into

solvent and most likely interacts with the T-cell receptor,

while P8 (K) forms a salt bridge with Q72. Also unlike

KLWG-9, threonine at P6 is oriented away from the

binding groove, toward solvent, providing a potential Tcr

interaction. The two prolines (P7 and P11) provide a

scaffold for further peptide-protein backbone interactions,

but may also interact with Tcr. P10-12 appear stacked such

Fig. 1 MS/MS spectrum of KLWG-12. y and b ions from peptide

fragmentation are labeled along with the mass of each fragment

Table 1 continued

Epitope SYFPEITHIscore Protein name

AKLLTIPQTLLNIS [14] KLLTIPQTLL 23 Zinc finger protein 493

AKLLTIPQTLLNIS [14] LTIPQTLLNI 23 Zinc finger protein 493

PVDTRAKVVLPPSLPRA [17] VLPPSLPRA 22 Uncharacterized protein KIAA1522

ALRGAGGGGVRDPGRLH [17] ALRGAGGGGV 25 Latent-transforming growth factor beta-binding protein

4 (LTBP-4)

For both T98g and CRL2610, multiple peptide fragments were identified from the same protein, so the longest parent fragment is listed along with

predicted HLA-binding epitopes contained within the parent sequence. SYFPEITHI scores below 20 were not considered


123

that threonine (P10) and serine (P12) form salt bridges with

K146 on HLA-A*0201, which could be responsible for the

increased relative affinity observed in T2 cell assays.

The molecular models of the docked FLWGLTPKV

(OPL1) and FLFGLTPKV (OPL2) peptides are similar

(Fig. 4c, d). P1 (F) interacts through a pi-pi aromatic

interaction with W167. The N-terminal NH2 group

H-bonds to the –OH group of Y7 and Y171, while F at P3

forms an aromatic interaction with Y159. Like KLWG-9,

P2 (L) and P9 (V) are buried in hydrophobic pockets of

the MHC. P7 (P) acts as a linker between the N- and

C-terminal portions of the peptide, and could interact with

Tcr. There are 6 H-bonds between the peptide backbone

and MHC-I clustered in the C-terminal area of the pep-

tide. All other H-bonding interactions occur with the

peptide backbone clustered at the C-terminal end of the

peptide. These modeling studies suggest that by simply

having more peptide-MHC contact, affinity may increase

if the peptide still fits into the binding groove. However,

it is important not to over-optimize a peptide for MHC

binding such that few residues are available to interact

with Tcr [32].

CTL generation and IFN-c ELISPOT assay

Human PBMC were cultured with repeated weekly stim-

ulations of either KLWG-9 or OPLs to determine whether

Fig. 2 PCR amplification of

HSD3B7 exons 2-4. a Oligo-

dT-primed cDNA from brain

tumor cell lines lane 1, CRL-

2610; lane 2 T98g; lane 3,

U251; lane 4, U-87. b Oligo-dT-

primed cDNA from primary

GBM line in lane 1, Ao2V4;

GBM tumor tissue lane 2, LB;

lane 3, LH; lane 4, NN; lane 5,

OJ; lane 6, TL. Lane 7 (SN) is

oligo-dT-primed cDNA from

normal brain tissue. In (a) and

(b), b-actin cDNA (189 bp) was

amplified across two exons as a

control for genomic

contamination. If genomic DNA

was present, a 600-bp band

would be observed

corresponding to an intron

between exons

Fig. 3 Detection of refolded HLA-A*0201 onto peptide beads.

a HLA-A*0201-positive control peptide; b HLA-A*0201-negative

control peptide; c KLWG-12 peptide; d KLWG-9 peptide

Table 2 Relative (IC50) and predicted affinities of OPL for HLA-A2

Computer

modeling

Peptide Sequencea SYFPEITHIb IC50

lMcTotal scored

KLWG-12 KLWGLTPKVTPS n/a 32 7.0

KLWG-9 KLWGLTPKV 30 110 3.0

OPL1 FLWGLTPKV 30 76 4.8

OPL2 FLFGLTPKV 30 17 13.4

a Standard single letter amino acid code and residues shown in bold represent

deviations from KLWG-9b SFXC-based total score and binding affinity for the variant-RT peptide (I1Y)

and the twenty-six peptides bound to the HLA-A*0201 crystal structurec Peptide epitope prediction values calculated using SYFPEITHId SFXC based total score and binding affinity for the Variant-RT peptide (I1Y)

and the twenty-six peptides bound to the HLA-A*0201 crystal structure


123

the peptides elicited a CTL response (Table 3). Because we

eluted KLWG-12 from GBM tumor cells, we hypothesized

that CTL raised against the 9-mer and the OPLs as pre-

sented by a functional APC should cross-react with tumors

expressing KLWG-12. To address this hypothesis, CTL

were raised on KLWG-9 peptide and the OPLs and evalu-

ated for recognition of native KLWG-12 peptide as well as

the cognate KLWG-9 in IFN-c ELISPOT assays. Each of

five donors responded to the peptides, but with different

profiles (Table 3). For donor ND51, KLWG-9 peptide

elicited nearly equivalent numbers of anti-KLWG-12 and

KLWG-9 CTL: 439 and 443 spots per 5 9 104 PBMC,

respectively. ND 51 responded only moderately to KLWG-

12 and KLWG-9 after stimulation with OPLs. In contrast,

CTL from donor ND78 raised on OPL1 peptide responded

better than wild-type and OPL2 peptide to the KLWG-12

and KLWG-9 target peptides with 97 and 105 spots per

5 9 104 PBMC, respectively. We were unable to raise

significant levels of CTL against the native peptides from

ND90 PBMC regardless of the peptide used for stimulation.

Donor ND93 CTL raised on OPL2 responded best to native

KLWG-9 peptide with 555 spots per 5 9 104 PBMC, but

curiously not to the 12-mer, as was observed with ND51.

CTL from donor ND94, raised on native KLWG-9 peptide,

responded weakly to KLWG-12 and cognate KLWG-9

peptides compared to other donors. All CTL except ND90

responded strongly to the same peptide they were stimu-

lated with cognate peptide (data not shown).

In view of the fact that KLWG-12 peptide was eluted

from a brain tumor cell line and demonstrated a stronger

Fig. 4 Peptide-HLA class I

interactions a KLWGLTPKV

(KLWG-9).

b KLWGLTPKVTPS (KLWG-

12), c FLWGLTPKV (OPL1),

d FLFGLTPKV (OPL2).

Peptide is shown as a stick

structure. HLA-A*0201 is

shown as a ribbon structure.

Hydrogen bonds between the

peptide and HLA-A*0201 are

shown by dotted lines and

labeled with the residue number

on the HLA molecule

Table 3 Peptide-reactive CTL elicited from normal donors responding to native KLWG-12 and KLWG-9 peptide epitopes as measured by

IFN-c ELISPOT

Donor ‘‘KLWG’’ target CTL raised on: KLWG-9 CTL raised on: FS-OPL1 CTL raised on: FS-OPL2

ND51 12-mer 439 (0.05) 52 (0.18) 19 (0.45)

ND51 9-mer 443 (0.02) 42 (0.21) 7 (0.29)

ND78 12-mer 88 (0.15) 97 (0.03) 25 (0.19)

ND78 9-mer 77 (0.29) 105 (0.03) 15 (0.23)

ND90 12-mer 9 (0.49) 11 (0.14) 8 (0.24)

ND90 9-mer 7 (0.39) 1 (0.39) 5 (0.33)

ND93 12-mer 68 (0.24) 49 (0.29) 27 (0.24)

ND93 9-mer 68 (0.15) 51 (0.16) 555 (0.01)

ND94 12-mer 30 (0.10) 5 (0.26) 0 (0.60)

ND94 9-mer 57 (0.37) 3 (0.36) 0 (0.17)

Table shows the number of spots from a 4th round of stimulation from CTL raised on OPL and tested on wild-type KLWG peptides. CTL were

tested against an irrelevant HIV peptide, and background spots were subtracted from totals. Values listed are IFN-c-secreting CTL from 5 9 104

PBMC per well using KLWG peptide-pulsed T2 cells as targets. Numbers in parentheses indicate coefficient of variation


123

affinity for HLA-A2 than the 9-mer, it was essential to

determine whether CTL could be generated against the

KLWG-12 peptide (Table 4). Therefore, normal human

PBMC were stimulated weekly with KLWG-12 peptide as

stated earlier. Donor ND117 CTL raised on KLWG-12

responded *35% more strongly to cognate KLWG-12

peptide than to KLWG-9 epitope with 279 and 181 spots,

respectively. Inhibition of CTL recognition of peptide-

loaded MHC class I molecule with W6/32 mAb demon-

strates the specificity and MHC dependency of the

anti-peptide CTL. ND117 CTL raised on KLWG-9 peptide

responded similarly to KLWG-9 and KLWG-12 peptides

with 204 and 209 spots, respectively. Blocking MHC class

I with W6/32 mAb again showed little response with 35

and 47 spots per 5 9 104 PBMC.

CTL cytotoxicity assay

At week 5 of CTL generation, we tested anti-peptide CTL

for the ability to lyse GBM tumor cells in an HLA-A2-

restricted manner in 51Cr release assays. At an effector-to-

target ratio of 100:1, 34% of CRL2610 GBM tumor cells

were lysed by ND 51 anti-KLWG-9 CTL (Fig. 5a). Pep-

tide-pulsed T2 cells were also lysed by ND51 CTL raised

on KLWG-9.

Anti-OPL1 CTL generated from donor ND51 were also

evaluated for the ability to lyse GBM tumor cells. At a

100:1 effector to target ratio, anti-OPL1 CTL lysed 33% of

CRL2610 tumor cells (Fig. 5b). Anti-OPL1 CTL from

ND51 were tested for their ability to lyse T2 cells pulsed

with the 12-mer peptide KLWG-12. At a 100:1 E:T ratio,

anti-OPL1 CTL showed 48% lysis, whereas the anti-

KLWG-9 CTL only showed 37% lysis.

Anti-KLWG-12 CTL generated from ND117 demon-

strated the ability to kill 41% of U-87 glioblastoma tumor

cells at an E:T ratio of 100:1, (Fig. 5c, solid diamonds).

MHC class I–dependent lysis of U-87 was inhibited with

the addition of blocking antibody W6/32 (Fig. 5c, solid

circle). In all cytotoxicity assays, NK activity was between

3 and 5%.

Discussion

Tumors expose their protein contents to CTL by presenting

them as peptides on cell surface MHC class I molecules.

One of the most direct methods for identifying tumor-

associated peptides is to elute them from MHC molecules

on the tumor cell surface and perform tandem mass spec-

trometry on the eluted peptides to determine their

sequences. Because tumor cells are genetically and trans-

criptionally unstable, it is possible that normal transcrip-

tional and translational machinery are dysregulated,

leading to translation of frameshifted and other abnormal

peptides. The abnormal proteins that result should be

ubiquitinylated and targeted to the proteasome where their

subsequent peptides would be sampled by MHC molecules

and presented on the tumor cell surface for T-cell recog-

nition [15–19, 33].

Mass spectrometric analyses of peptides acid-eluted

from GBM tumor cell lines, T98g and CRL2610, resulted

in identification of several potential antigenic peptides. Of

particular interest was a 12-amino acid frameshift peptide

derived from the HSD3B7 gene. It was predicted to have a

very high SYFPEITHI binding score. This peptide was not

found in acid elutions from CRL2610 or PBMC from two

unrelated healthy donors using the same LC–MS/MS

methods. Interestingly, frameshifts from the zinc finger

protein 493 (ZNF493) gene were common to both T98g

and CRL2610, suggesting that this gene may be predis-

posed to transcriptional or translational infidelity. ZNF493

has 3 known variants; isoform 1 lacks 2 exons and utilizes

a downstream start codon, while isoform 2 includes an

exon from a putative 30 untranslated region.

Neither PBMC nor normal brain tissue is available from

the same patient for which the T98g cell line was derived.

However, it is unlikely that this frameshift results from a

unique, patient-specific mutation for 2 reasons. First, PCR

amplification of the RNA region flanking the KLWG

peptide resulted in a splice variant present in 5 GBM

tumors, one primary cell line and all 4 established GBM

cell lines, but not in normal brain tissue (Fig. 2). Second,

Table 4 Peptide-reactive CTL elicited from ND117 responding to native KLWG-12 and KLWG-9 peptide epitopes as measured by IFN-cELISPOT

Donor KLWG ‘‘Target’’ CTL Raised on KLWG-9 CTL Raised on KLWG-12

ND117 12-mer 209 (0.11) 279 (0.09)

ND117 12-mer ? W6/32 35 (0.4) 34 (0.11)

ND117 9-mer 204 (0.11) 181 (0.21)

ND117 9-mer ? W6/32 47 (0.12) 33 (0.12)

Table shows the number of spots from a 4th round of stimulation from CTL generated on KLWG-12, KLWG-9, and tested on wild-type KLWG

peptides with or without MHC class I blocking antibody W6/32. CTL were tested against an irrelevant HIV peptide, and background spots were

subtracted from totals. Values listed are IFN-c-secreting CTL from 5 9 104 PBMC per well using KLWG peptide-pulsed T2 cells as targets.

Numbers in parentheses indicate coefficient of variation


123

KLWG-12 was identified from T98g eluates, and CTL

raised against the peptide kill CRL2610 and U-87, unre-

lated GBM cell lines. It is clear from Fig. 2 that two splice

variants exist for HSD3B7 in tumor cells and tissue, but not

in normal brain tissue. According to DNA sequence anal-

ysis, the 438-bp PCR product contains exons 2, 3, and 4,

while the 282-bp band contains exons 2 and 4 spliced

together. Unfortunately, we did not find evidence for mis-

splicing in the 282-bp band, even after cloning the PCR

products and sequencing nearly 100 clones. Alternative and

aberrant splicing is well-known in cancer [34–36]. Epi-

dermal growth factor receptor VIII is one well-known

example of a prevalent splice variant in GBM that is also a

target for immunotherapy [37].

It is possible that the junction between exons 2 and 4 is

problematic for translational machinery resulting in ribo-

somal frameshifting, stalling, or slippage [38–40]. These

mechanisms may lead to intermittent expression of the

KLWG peptide such that some transcripts are in frame and

some are translated out of frame. We are currently

addressing this complex question.

The 12-mer peptide contains a 9-mer epitope embedded

in the parent 12-mer that was predicted to bind HLA-

A*0201 by computer algorithms. Interestingly, both the

12-mer and the 9-mer bind to recombinant HLA-A*0201 in

a refolding assay, confirming the hypothesis that KLWG-

12 binds to HLA-A*0201. Surprisingly, the 12-mer bound

to cell surface HLA-A*0201 with a relative affinity three

times stronger than the 9-mer and at least twice as strongly

as OPL1 which contains a preferred phenylalanine (F) at

P1 (Table 2). Although there are no crystal structures

greater than 10 amino acids bound to an HLA molecule,

the molecular modeling scores (higher values indicate

stronger binding, unlike IC50 values) are in agreement with

the cell-based relative affinity measurement, IC50.

Although KLWG-12 demonstrated the strongest relative

affinity to HLA-A*0201, stimulation of normal donor

PBMC with KLWG-9 elicited the strongest responses

among 5 normal donors. CTL were raised on KLWG-9,

KLWG-12, and the two 9-mer OPLs. The rationale for

raising CTL against 9-mers is that functional APCs are

more likely to process and present 9-mers, while tumor

cells may have relaxed antigen presentation mechanisms

such that they present longer peptides. However, the rela-

tive affinity for KLWG-12 peptide for HLA-A2 was 3-fold

stronger than the 9-mer, so CTL were also generated

against KLWG-12. Starting at week five of stimulation,

CTL were assayed by ELISPOT for reactivity only to

native KLWG-12 and KLWG-9 peptides (Table 3).

Responses from different donors to the peptides were

variable, as observed in previous studies [30].

To discuss these findings in a vaccine setting, if donor

ND78 had a glioblastoma expressing KLWG-12 and was

immunized with native KLWG-9, he/she would not

respond as well as if he/she would have been immunized

with OPL1. Similarly, one might argue that donor ND93

should only be immunized with OPL2 because other pep-

tides only sub-optimally stimulated anti-KLWG CTL. CTL

from donor ND51 responded equally well to native

KLWG-12 and KLWG-9, but were marginally stimulated

by OPL1 and only weakly stimulated by OPL2. Although

donor ND90 PBMC were stimulated by a positive control

influenza peptide (data not shown), we were not able to

0

10

20

30

40

50

60

70

80

90

100

100:1 50:1 25:1 12.5:1

% S

peci

fic

Lys

is

E:T Ratio

(A) CTL raised on KLWG-9

(B) CTL raised on OPL1

(C) CTL raised on KLWG -12

Fig. 5 Lysis of target and GBM tumor cells. CTL at indicated E:T

ratios were incubated with 51Cr-labeled target cells (5 9 103 cells/

well), and 51Cr release was measured after 8 h. a ND 51 CTL raised

on native KLWG-9 peptide b ND 51 CTL raised on OPL1 peptide.

GBM CRL2610 (solid circle). T2 cells pulsed 10 lg/ml with KLWG-

12 (solid diamond) and KLWG-9 (solid square). K562 (open square)

served as NK sensitive control. c ND117 CTL raised on KLWG-12

peptide. U-87 (solid diamond). U-87 with MHC class I blocking

antibody (solid circle). K562 (open square) served as NK sensitive

control


123

elicit CTL from ND90 with any of the KLWG or OPL

peptides. This might suggest that ND90 may have a hole in

his/her T-cell receptor repertoire, or perhaps there is an

OPL not examined in this study that would stimulate CTL

from ND90 to respond to native KLWG peptides. Alter-

natively, donor ND90 may be tolerant to KLWG and

related peptides. Similar to ND90, the response of ND94 to

KLWG-9 was weak, and non-existent to both OPLs.

Overall, ELISPOT data strongly suggest that the KLWG

12-mers and perhaps 9-mers are expressed on the tumor

cell surface for CTL recognition. The ELISPOT data also

parallel to what is observed in patients after vaccination

with tumor-derived peptides; some patients respond, while

some do not. Further studies may indicate whether patients

whose tumors express KLWG peptide possess anti-KLWG

CTL. It is worth speculating that one might increase the

frequency of response among patients if multiple OPL

were used in the vaccine. These findings are also further

evidence that peptides longer than the canonical 8-10 res-

idues can be presented by MHC molecules if they contain a

core HLA-binding epitope [4–6]. To prove that KLWG

peptide recognition was MHC dependent, W6/32 antibody

was used to block ND117 CTL recognition of target cells

in ELISPOT (Table 4).

Cytotoxicity assays were performed using CTL from

five of the six donors raised on each peptide (native

KLWG-9. -12, OPL1 and 2) to determine whether their

CTL would kill HLA-A2? glioblastoma cells. Only CTL

from donor ND51 and ND117 demonstrated the ability to

kill GBM tumor cells. Interestingly, ND51 killed CRL2610

HLA-A2-positive glioblastoma multiforme tumor cells

(Fig. 5), while ND117 killed U-87 tumor cells. Once again

it is not surprising that different donors respond differently

to different tumor cells [30].

We observed similar cytolytic profiles between CTL

raised on native KLWG-9 peptide and CTL raised on

OPL1. Interestingly, KLWG-9 and OPL CTL did not kill

the tumor cell line, T98g, in which the peptide was found

(data not shown). This is not unusual. We have previously

reported that the T98g is more difficult to kill than other

GBM tumor cell lines due to TGF-b secretion [30] and

perhaps other inhibitory cytokines [41, 42]. CTL killing

results indicate that CRL2610 and U-87 tumor cells

express KLWG-9 and/or KLWG-12-mer. This suggests

that CTL are more sensitive to the presence of peptide

than our mass spectrometric detection methods; biology is

more sensitive than technology. Although lysis of GBM

from donor ND51 CTL raised on native peptide or OPL1

were similar, CTL raised on KLWG-9 were most

responsive in ELISPOT (439 spots), but CTL from the

same donor raised on OPL1 peptide showed only 52 spots.

This result is in line with other findings showing a lack of

correlation between cytolytic activity and IFN-c ELISPOT

numbers [43]. This result suggests that the differences in

cytolytic activity may be related more to inherent differ-

ences among donors than to the peptide used for

stimulation.

An additional observation is that the large CTL:target

ratios results in a small variation in killing. It is possible

that some of the CRL2610 and U-87 target cells are het-

erogeneous and do not express high levels of MHC class I.

Alternatively, some of the cells may not express the

KLWG frameshift peptide due to antigenic modulation or

loss of antigen. Either or both of these possibilities would

result in low initial lysis at 100:1 in which most of the cells

expressing peptide-MHC are killed, but those that do not

are not killed.

The results presented here show that by examining the

MHC-binding peptidome of tumor cells, new peptide

antigens can be discovered as a result of genomic, tran-

scriptional, and translational instability. The results also

support rational design and optimization of peptides

derived from tumor-specific antigens for development of

peptides with increased binding affinity for HLA molecules

such that they stimulate a broad T-cell response to an

individual antigenic epitope so that as many anti-epitope

CTL will be stimulated as possible. Relative affinity

measurements of peptides with HLA-A*0201 can be cou-

pled with functional assays to measure peptide stimulation

of CTL. This is especially important given that different

donors respond differentially to native or optimized pep-

tides from the same epitope, suggesting that immunization

with multiple peptides for the same epitope will elicit CTL

from more individuals than immunization with one epitope.

However, it is well established that tumors can down-

modulate individual peptides to escape detection by the

immune system. Our findings do not address this problem,

but instead further substantiate the importance of immu-

nizing individuals not only with multiple tumor antigens

but also with multiple peptides from the same epitope to

elicit the broadest response within the same patient and

among different patients. The results presented here may

explain, in part, the variability in individual patient

responses to immunotherapy.

Acknowledgments We would like to thank HoJoon Lee, M.S. for

the frameshift database.

References

1. Hou LC, Veeravagu A, Hsu AR, Tse VC (2006) Recurrent

glioblastoma multiforme: a review of natural history and man-

agement options. Neurosurg Focus 20:E5

2. Laws ER, Parney IF, Huang W et al (2003) Survival following

surgery and prognostic factors for recently diagnosed malignant

glioma: data from the Glioma Outcomes Project. J Neurosurg

99:467–473


123

3. Novellino L, Castelli C, Parmiani G (2005) A listing of human

tumor antigens recognized by T cells: March 2004 update. Cancer

Immunol Immunother 54:187–207

4. Antwi K, Hanavan PD, Myers CE, Ruiz YW, Thompson EJ, Lake

DF (2009) Proteomic identification of an MHC-binding pepti-

dome from pancreas and breast cancer cell lines. Mol Immunol

46:2931–2937

5. Stryhn A, Pedersen LO, Holm A, Buus S (2000) Longer peptide

can be accommodated in the MHC class I binding site by a

protrusion mechanism. Eur J Immunol 30:3089–3099

6. Gebreselassie D, Spiegel H, Vukmanovic S (2006) Sampling of

major histocompatibility complex class I-associated peptidome

suggests relatively looser global association of HLA-B*5101

with peptides. Hum Immunol 67:894–906

7. Falk K, Rotzschke O, Stevanovic S, Jung G, Rammensee HG

(1991) Allele-specific motifs revealed by sequencing of self-

peptides eluted from MHC molecules. Nature 351:290–296

8. Stevanovic S, Schild H (1999) Quantitative aspects of T cell

activation–peptide generation and editing by MHC class I mol-

ecules. Semin Immunol 11:375–384

9. Hunt DF, Henderson RA, Shabanowitz J et al (1992) Charac-

terization of peptides bound to the class I MHC molecule HLA-

A2.1 by mass spectrometry. Science 255:1261–1263

10. Kast WM, Brandt RM, Sidney J et al (1994) Role of HLA-A

motifs in identification of potential CTL epitopes in human

papillomavirus type 16 E6 and E7 proteins. J Immunol

152:3904–3912

11. Parker KC, Bednarek MA, Hull LK et al (1992) Sequence motifs

important for peptide binding to the human MHC class I mole-

cule, HLA-A2. J Immunol 149:3580–3587

12. Pogue RR, Eron J, Frelinger JA, Matsui M (1995) Amino-ter-

minal alteration of the HLA-A*0201-restricted human immuno-

deficiency virus pol peptide increases complex stability and in

vitro immunogenicity. Proc Natl Acad Sci USA 92:8166–8170

13. Ruppert J, Sidney J, Celis E, Kubo RT, Grey HM, Sette A (1993)

Prominent role of secondary anchor residues in peptide binding to

HLA-A2.1 molecules. Cell 74:929–937

14. Smith MH, Nuara AA, Egen JG, Sirjani DB, Lam KS, Grimes WJ

(1998) Baculoviral expressed HLA class I heavy chains used to

screen a synthetic peptide library for allele-specific peptide

binding motifs. Mol Immunol 35:1033–1043

15. Linnebacher M, Gebert J, Rudy W et al (2001) Frameshift pep-

tide-derived T-cell epitopes: a source of novel tumor-specific

antigens. Int J Cancer 93:6–11

16. Saeterdal I, Bjorheim J, Lislerud K et al (2001) Frameshift-

mutation-derived peptides as tumor-specific antigens in inherited

and spontaneous colorectal cancer. Proc Natl Acad Sci USA

98:13255–13260

17. Schwitalle Y, Linnebacher M, Ripberger E, Gebert J, von Knebel

Doeberitz M (2004) Immunogenic peptides generated by frame-

shift mutations in DNA mismatch repair-deficient cancer cells.

Cancer Immun 4:14

18. Townsend A, Ohlen C, Rogers M, Edwards J, Mukherjee S,

Bastin J (1994) Source of unique tumour antigens. Nature

371:662

19. Elliott T, Bodmer H, Townsend A (1996) Recognition of out-of-

frame major histocompatibility complex class I-restricted epi-

topes in vivo. Eur J Immunol 26:1175–1179

20. Cheng JB, Jacquemin E, Gerhardt M et al (2003) Molecular

genetics of 3beta-hydroxy-Delta5–C27-steroid oxidoreductase

deficiency in 16 patients with loss of bile acid synthesis and liver

disease. J Clin Endocrinol Metab 88:1833–1841

21. van der Burg SH, Ras E, Drijfhout JW et al (1995) An HLA class

I peptide-binding assay based on competition for binding to class

I molecules on intact human B cells. Identification of conserved

HIV-1 polymerase peptides binding to HLA-A*0301. Hum

Immunol 44:189–198

22. Korbie DJ, Mattick JS (2008) Touchdown PCR for increased

specificity and sensitivity in PCR amplification. Nat Protocols

3:1452–1456

23. Smith MH, Lam KS, Hersh EM, Lebl M, Grimes WJ (1994)

Peptide sequences binding to MHC class I proteins. Mol Immunol

31:1431–1437

24. Lam KS, Lebl M, Krchnak V (1997) The ‘‘One-Bead-One-Com-

pound’’ combinatorial library method. Chem Rev 97:411–448

25. Lam KS, Salmon SE, Hersh EM, Hruby VJ, Kazmierski WM,

Knapp RJ (1991) A new type of synthetic peptide library for

identifying ligand-binding activity. Nature 354:82–84

26. Dionne SO, Myers CE, Smith MH, Lake DF (2004) Her-2/neu

altered peptide ligand-induced CTL responses: implications for

peptides with increased HLA affinity and T-cell-receptor inter-

action. Cancer Immunol Immunother 53:307–314

27. Parkhurst MR, Salgaller ML, Southwood S et al (1996) Improved

induction of melanoma-reactive CTL with peptides from the

melanoma antigen gp100 modified at HLA-A*0201-binding

residues. J Immunol 157:2539–2548

28. Kumar N, Mohanty D (2007) MODPROPEP: a program for

knowledge-based modeling of protein-peptide complexes.

Nucleic Acids Res 35:W549–W555

29. Rammensee H, Bachmann J, Emmerich NP, Bachor OA, Steva-

novic S (1999) SYFPEITHI: database for MHC ligands and

peptide motifs. Immunogenetics 50:213–219

30. Myers CE, Dionne SO, Shakalya K, Mahadevan D, Smith MH,

Lake DF (2008) Variation in cytotoxic T-lymphocyte responses

to peptides derived from tyrosinase-related protein-2. Hum

Immunol 69:24–31

31. Dionne SO, Smith MH, Marincola FM, Lake DF (2003) Func-

tional characterization of CTL against gp100 altered peptide

ligands. Cancer Immunol Immunother 52:199–206

32. Dionne SO, Smith MH, Marincola FM, Lake DF (2001) Antigen

presentation of a modified tumor-derived peptide by tumor

infiltrating lymphocytes. Cell Immunol 214:139–144

33. Ishikawa T, Fujita T, Suzuki Y et al (2003) Tumor-specific

immunological recognition of frameshift-mutated peptides in

colon cancer with microsatellite instability. Cancer Res

63:5564–5572

34. Belicha-Villanueva A, Blickwedehl J, McEvoy S, Golding M,

Gollnick SO, Bangia N (2010) What is the role of alternate

splicing in antigen presentation by major histocompatibility

complex class I molecules? Immunol Res 46:32–44

35. Ward AJ, Cooper TA (2010) The pathobiology of splicing.

J Pathol 220:152–163

36. Kalnina Z, Zayakin P, Silina K, Line A (2005) Alterations of pre-

mRNA splicing in cancer. Genes Chromosomes Cancer

42:342–357

37. Gan HK, Kaye AH, Luwor RB (2009) The EGFRvIII variant in

glioblastoma multiforme. J Clin Neurosci 16:748–754

38. Farabaugh PJ (1996) Programmed translational frameshifting.

Annu Rev Genet 30:507–528

39. Fernandez J, Yaman I, Huang C et al (2005) Ribosome stalling

regulates IRES-mediated translation in eukaryotes, a parallel to

prokaryotic attenuation. Mol Cell 17:405–416

40. Hansen TM, Reihani SN, Oddershede LB, Sorensen MA (2007)

Correlation between mechanical strength of messenger RNA

pseudoknots and ribosomal frameshifting. Proc Natl Acad Sci

USA 104:5830–5835

41. Bodmer S, Strommer K, Frei K et al (1989) Immunosuppression

and transforming growth factor-beta in glioblastoma. Preferential

production of transforming growth factor-beta 2. J Immunol

143:3222–3229


123

42. Finke J, Ferrone S, Frey A, Mufson A, Ochoa A (1999) Where

have all the T cells gone? Mechanisms of immune evasion by

tumors. Immunol Today 20:158–160

43. van Besouw NM, Zuijderwijk JM, de Kuiper P, Ijzermans JN,

Weimar W, van der Mast BJ (2005) The granzyme B and

interferon-gamma enzyme-linked immunospot assay as alterna-

tives for cytotoxic T-lymphocyte precursor frequency after renal

transplantation. Transplantation 79:1062–1066


123

CTL recognition of a novel HLA-A*0201-binding peptide derived from glioblastoma multiforme tumor cells

Documents