Phosphorylated self-peptides alter human leukocyteantigen class I-restricted antigen presentationand generate tumor-specific epitopesJan Petersena,1, Stephanie J. Wurzbacherb,1, Nicholas A. Williamsonb, Sri H. Ramarathinamb, Hugh H. Reida,Ashish K. N. Nairb, Anne Y. Zhaob, Roza Nastovskab, Geordie Rudgeb, Jamie Rossjohna,2, and Anthony W. Purcellb,2

aProtein Crystallography Unit, Department of Biochemistry and Molecular Biology, School of Biomedical Sciences, Monash University, Victoria 3800,Australia; and bDepartment of Biochemistry and Molecular Biology and Bio21 Molecular Science and Biotechnology Institute, Universityof Melbourne, Victoria 3010, Australia

Communicated by Peter Doherty, University of Melbourne, Victoria, Australia, December 22, 2008 (received for review November 30, 2008)

Human leukocyte antigen (HLA) class I molecules present a varietyof posttranslationally modified epitopes at the cell surface, al-though the consequences of such presentation remain largelyunclear. Phosphorylation plays a critical cellular role, and deregu-lation in phosphate metabolism is associated with disease, includ-ing autoimmunity and tumor immunity. We have solved thehigh-resolution structures of 3 HLA A2-restricted phosphopeptidesassociated with tumor immunity and compared them with thestructures of their nonphosphorylated counterparts. Phosphoryla-tion of the epitope was observed to affect the structure andmobility of the bound epitope. In addition, the phosphoamino acidstabilized the HLA peptide complex in an epitope-specific mannerand was observed to exhibit discrete flexibility within the antigen-binding cleft. Collectively, our data suggest that phosphorylationgenerates neoepitopes that represent demanding targets for T-cellreceptor ligation. These findings provide insights into the mode ofphosphopeptide presentation by HLA as well as providing a plat-form for the rational design of a generation of posttranslationallymodified tumor vaccines.

antigen presentation � HLA � phosphopeptide � T cells �X-ray crystallography

Phosphorylation plays a critical role in cellular signaling, andchanges in phosphate metabolism are associated with virtu-

ally all disease states. The immune system has evolved to surveychanges in phosphorylation through the action of both innateand adaptive effector pathways that include specific recognitionof phosphoantigens. For example, Toll-like receptor (TLR) 4and TLR9 perceive various forms of phosphoantigens (1, 2).Moreover, a subset of �� T cells recognizes pyrophosphomo-noesters that are found in various microbial pathogens (3).Natural killer (NK) cell recognition of phosphoantigens has alsorevealed that phosphorylation of human leukocyte antigen(HLA) Cw4-bound peptide antigens reduced inhibitory signalsmediated via killer Ig receptors and led to enhanced NK cellcytolysis (4). Phosphoantigens are also recognized by the adap-tive immune system. Recognition of phosphoantigens by anti-bodies is very well documented (5), and phosphorylated autoan-tigens are implicated in human autoimmune disorders, such asprimary Sjogren’s syndrome and lupus (6, 7). Phosphoantigensurveillance by T cells has been observed in major histocom-patibility complex (MHC) class I- and class II-restricted antigenpresentation (8–10). In addition, HLA A2-restricted tumor-specific phosphopeptides are immunogenic, and cytotoxic Tlymphocytes (CTLs) that distinguish between phosphorylatedand native peptides can be generated in HLA A2 transgenicmice (9).

The exquisite sensitivity of CTLs toward subtle changes inpeptides presented on the cell surface allows discrimination ofinfected cells or cells undergoing malignant transformation.

Cancer immunotherapy has focused on the identification oftumor-associated antigens that are expressed exclusively bycancer cells. These antigens fall into 3 broad classes: (i) cancerantigens, such as testis and other embryonic or developmentalantigens that are not normally expressed in adult tissues but areexpressed in a broad range of tumors (11); (ii) neoantigensgenerated by mutation in key regulator molecules, such as p53(12) or aberrant posttranslational modification of proteins (13);and (iii) viral antigens associated with cancer, such as Epstein-Barr virus antigens (14). Many of these antigens are not ex-pressed on the surface of tumor cells, and therefore are notdirectly accessible to antibodies. Thus, because of the ability ofCTLs to survey intracellular protein expression, vaccines that arecapable of eliciting such responses represent an attractive optionfor cancer immunotherapy.

To study the intrinsic link between the deregulated signalingcascade present in many cancers and the ability of antigenprocessing to alert CTLs to such molecular events, we haveinvestigated the structural and biophysical properties and struc-tures of 3 HLA A2 phosphopeptide complexes derived from celldivision cycle (CDC) 25b, �-catenin, and insulin receptor sub-strate (IRS) 2 and have compared them with the structures oftheir nonphosphorylated counterparts. The presentation ofHLA class I-restricted phosphorylated epitopes and the impli-cations for altered self are discussed.

Results and DiscussionStructures of HLA A2 Bound to Phospho- and Native-Peptide Epitopes.To gain insight into the mode of phosphopeptide presentation,HLA A2 was expressed and refolded in the presence of 3phosphopeptides. Two of these peptides, a nonamer anddecamer, were phosphorylated at the P4 position IRS21097RVApSPTSGV1105, �-catenin 30YLDpSGIHSGA39,whereas the other nonamer was phosphorylated at the P5position CDC25b residues 38–46 38GLLGpSPVRA46. Thesephosphoserine-containing peptides were chosen based on theirnatural antigen presentation on multiple HLA A2� tumor celllines and their immunogenicity in HLA A2 transgenic mice (9).

Author contributions: J.R. and A.W.P. designed research; J.P., S.J.W., N.A.W., H.H.R.,A.K.N.N., A.Y.Z., and R.N. performed research; S.H.R. and G.R. contributed new reagents/analytical tools; J.P. and S.J.W. analyzed data; and J.P., S.J.W., J.R., and A.W.P. wrote thepaper.

The authors declare no conflict of interest.

Data deposition: The atomic coordinates have been deposited in Protein Data Bank,www.pdb.org (PDB ID codes 3FQN, 3FQR, 3FQT, 3FQU, 3FQW, and 3FQX).

1J.P. and S.J.W. contributed equally to this work.

2To whom correspondence may be addressed. E-mail: jamie.rossjohn@med.monash.edu.auor apurcell@unimelb.edu.au.

This article contains supporting information online at www.pnas.org/cgi/content/full/0812901106/DCSupplemental.

© 2009 by The National Academy of Sciences of the USA

2776–2781 � PNAS � February 24, 2009 � vol. 106 � no. 8 www.pnas.org�cgi�doi�10.1073�pnas.0812901106

Dow

nloa

ded

by g

uest

on

Oct

ober

7, 2

020

The 3 HLA A2 epitopes were subsequently crystallized, andtheir respective structures were solved and refined to a resolutionof 1.8 Å or better (Table 1). In addition, to gain insight into howthe incorporation of the phospho-moiety influenced the pHLAA2 structure, we determined the structures of the nonphospho-rylated counterparts of these peptides bound to HLA A2 to aresolution of 1.93 Å or better (Table 1). With the exception ofthe HLA A2YLDSGIHSGA, in which the central region of theepitope demonstrated high mobility (see below), the mode ofbinding of the peptides was unambiguous (Fig. 1).

All 6 pHLA A2 structures determined were crystallized underthe same conditions, in the same space group and unit celldimensions (Table 1). In addition, an alignment of the antigen-binding cleft (residues 1–185) of HLA A2 indicated no signifi-cant structural rearrangement in the corresponding phospho-and nonphospho-structures (rmsd: CDC25b/CDC25b-phospho,HLA A2 � 0.079 Å, peptide � 0.27 Å; IRS2/IRS2-phospho,HLA A2 � 0.13 Å, peptide � 0.09 Å; �-catenin/�-catenin-phospho, HLA A2 � 0.11 Å, peptide � 0.45 Å). In all cases, thephosphate group is prominently surface-exposed and contrib-utes to increased electronegativity at the candidate T-cell re-ceptor (TCR) binding site (Figs. 2 and 3).

CDC25b. The CDC25b peptide (GLLGSPVRA) bound to HLAA2 in a linear and extended manner with a central bulge aroundthe P4-P5 position (Fig. 1 A). P2-Leu and P9-Ala are the mainanchor residues, whereas P3-Leu also pointed down toward theantigen-binding cleft and contributes to peptide binding. P5-Serand P8-Arg interact via a water-mediated H-bond, and bothproject upward, comprising potential TCR contact sites, withP5-Ser leaning toward the �1-helix, and forming van der Waalsinteractions with Ala 69 of HLA A2 (Fig. 2 A). There are noconformational changes in the HLA A2 heavy chain associatedwith the accommodation of the phospho-moiety in the HLAA2GLLGpSPVRA complex (Fig. 1D). However, the peptide under-

goes local conformational changes around the site of phosphor-ylation at P5 to avoid steric clashes with Ala 69 and Thr 73 of theHLA A2 heavy chain (Fig. 2 B and C). This results in the peptidepushing away from the �1-helix toward center of the antigen-binding cleft, resulting in a shift of 2 Å in the C� positionbetween P5 and P5-phosphoSer and also a change in theconformation of P3-Leu. Moreover, the phosphate group isobserved in 2 discrete conformations (Fig. 2B), indicative ofdiscrete mobility in this moiety [Table S1]. One conformer formsa salt bridge with P8-Arg, and both conformers interact with thepeptide backbone through a water-mediated H-bond to P6-Pro,which appear to be the only interactions the phosphate headgroup makes (Fig. 2 A and B). In this instance, these pHLAcomplexes define a case of altered self in which the peptideantigen demonstrates significantly altered conformation. Thesedata demonstrate the alteration of a self-pHLA by a posttrans-lational modification.

IRS2. The IRS2 peptide (RVASPTSGV) also bound to HLA A2in an extended manner with P2-Val and P9-Val as main anchorresidues (Fig. 1B). In addition, P3-Ala and P6-Thr point down-ward into the antigen-binding cleft. P1-Arg, P4-Ser, P5-Pro, andP7-Ser are solvent-exposed and potential TCR contact sites,although P4-Ser is not involved in any significant interactions.The phosphorylated IRS2 peptide is accommodated within theHLA A2 binding cleft with little change when compared withthe nonphosphorylated counterpart (Fig. 1E). The differences inthe phosphorylated and nonphosphorylated complexes residepredominantly in the addition of an electronegative charge andsmall changes in the conformation of Lys 66 and Arg 65 (Fig. 2D–F). As observed for the CDC25b pHLA complex, the phos-phate group at P4 of the IRS2 peptide was observed in 2conformations, again reflecting flexibility in the phospho-moiety. One conformer interacts with Lys 66 on the �1-helix andwith Gln 155 on the �2-helix through a water-mediated H-bond

Table 1. Data collection and refinement statistics

IRS2nonphospho

IRS2phospho

�-cateninnonphospho

�-cateninphospho

CDC25bnonphospho

CDC25bphospho

Resolution, Å 23.7–1.93 24.0–1.70 23.7–1.65 24.3–1.70 24.9–1.80 30–1.80Space group P212121 P212121 P212121 P212121 P212121 P212121

Cell dimensions, Å (a, b, c) 59.9879.38111.29

59.8579.68111.00

59.7279.71110.87

59.6879.71111.56

59.6879.84110.21

59.9180.04

110.37Total no. observations 38,009 56,258 63,777 56,745 49,135 49,810Multiplicity 4.7 5.1 6.3 7.1 3.5 4.9Data completeness, % 93.07 (75.9) 95.25 (73.9) 98.7 (88.8) 95.8 (76.3) 99.19 (97.8) 99.8 (100)I/�I 22.1 (3.7) 26.1 (3.9) 27.3 (2.5) 26.9 (4.3) 18.6 (2.8) 23.9 (3.3)Rmerge*, % 5.6 (32.8) 4.4 (31.8) 5.0 (42.9) 5.1 (30.4) 5.0 (45.0) 5.2 (46.6)Rfactor

†, % 16.85 17.10 17.92 16.85 17.53 17.04Rfree

‡, % 19.63 19.61 19.95 18.86 20.34 20.18rmsd from idealityBond lengths, Å 0.006 0.008 0.006 0.005 0.005 0.007Bond angles, ° 1.022 1.115 1.058 0.998 0.967 1.142

Ramachandran angles, %Favored 97.61 98.13 97.84 97.61 98.14 98.40Allowed 2.39 1.87 2.16 2.39 1.86 1.60Outliers — — — — — —

B-factorsPeptide 28.7 29.0 28.0 26.4 28.0 38.3Protein 27.3 28.5 27.3 24.4 29.4 27.6

Water ions (Cd, Co, Mg), glycerol 36.5 41.3 40.3 38.0 40.1 38.7

*Rmerge � � �Ihkl��Ihkl��/�Ihkl.†Rfactor � �hkl � �Fo���Fc� �/�hkl �Fo� for all data except for 5%, which was used for the ‡Rfree calculation. Numbers in parentheses refer to statistics in the highestresolution bin.

Petersen et al. PNAS � February 24, 2009 � vol. 106 � no. 8 � 2777

IMM

UN

OLO

GY

Dow

nloa

ded

by g

uest

on

Oct

ober

7, 2

020

(Table S1). The second conformer interacts with the Lys 66 andArg 65 of HLA A2. Thus, the phospho-moiety forms stabilizingcontacts with the antigen-binding cleft without disrupting thepeptide conformation.

A recent structural study of phosphopeptide/HLA A2 com-plexes suggested a common binding motif for the P4-phosphoSermoiety of peptides, with a positively charged N-terminal residue

(15). The binding mode of the phosphorylated IRS2 peptide onlypartly follows this motif, however, because the salt bridge toP1-Arg is absent in the HLA A2RVApSPTSGV complex. This canbe attributed to the extended binding geometry of P5-Pro andP6-Thr, which prevents the movement of P4-Ser toward P1 intothe position observed by Mohammed et al. (15). Instead, P1-Argand Lys 66 adopt side-chain conformations that place the centerof the positively charged region somewhat closer to the phos-phate moiety.

Fig. 1. Peptide conformations within the antigen-binding cleft. CDC25b (A),IRS2 (B), �-catenin (C), CDC25b-phospho (D), IRS2-phospho (E), and �-catenin–phospho (F). Blue mesh indicates unbiased 2Fo-Fc maps contoured at 1�.Yellow indicates nonphosphorylated peptides. Green indicates phosphory-lated peptides. The bound peptide is shown from a side-on view with the�2-helix removed for clarity.

A B C

D E F

G H I

Fig. 2. Interactions of the phosphorylation site in nonphosphopeptide andphosphopeptide HLA A2 complexes. Accommodation of the phosphate moi-ety by HLA A2 is accompanied by changed interactions in the complex, addingto the differential presentation of altered self. Stick representation of pep-tides and of heavy-chain side chains that interact with the phosphorylationsite. Yellow indicates nonphospho-pHLA A2 complexes. Green indicatesphospho-pHLA A2 complexes. (A–C) Phosphorylation of P5-Ser in CDC25bleads to an altered peptide conformation attributable to steric constraints.(D–F) Phosphorylation of P4-Ser in IRS2 gives rise to numerous interactions andsubtly alters the conformation of Arg 65 and Lys 66. (G–I) Phosphorylation ofP4-Ser in �-catenin stabilizes the mobile peptide residues P3 to P6.

Fig. 3. Altered surface potential for TCR recognition. Surface representationof the HLA A2 with bound peptides. CDC25b (A), IRS2 (B), CDC25b-phospho(C), and IRS2-phospho (D). Gray indicates �-chain, with putative TCR contactresidues in purple, based on the structure of the A6/HLA A2-Tax complexstructure (25). The arrows indicate the peptide phosphorylation sites. Thenegatively charged phosphate groups are located within the area of a typicalTCR footprint and are likely to dominate TCR discrimination. Electrostaticpotentials (blue, positive; red, negative) were calculated with APBS (22). Thephosphoserine residues were assumed to carry 2 negative charges.

2778 � www.pnas.org�cgi�doi�10.1073�pnas.0812901106 Petersen et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

7, 2

020

�-Catenin. The native form of this peptide (YLDSGIHSGA)demonstrates marked flexibility in complex with HLA A2, withno electron density observed for positions 5 and 6 (Fig. 1C). Inaddition, P3-Asp, P4-Ser, and P7-His exhibit f lexibility. Themobility of the central region of the �-catenin peptide is notattributable to poor crystallographic data, because the electrondensity surrounding this region is excellent and, moreover, thestructure is at very high resolution (1.65 Å) and well refined[Rfree � 19.95%]. There is also a degree of mobility of theresidues in the floor of the antigen-binding cleft (e.g., Tyr 99, Arg97) as a result of the peptide flexibility (data not shown).Although the peptide is highly flexible, the primary anchor(P2-Leu and P10-Ala) residues are well defined.

The mobility of the nonphosphorylated pHLA is markedlyreduced in the phosphorylated pHLA structure, and the entireepitope is clearly visible (Fig. 1F). Consistent with the other 2phosphorylated complexes, the phosphate moiety is observed in2 different conformers, which indicates a general theme of amobile phosphate group (Fig. 2 G–I). In the HLAA2YLDpSGIHSGA complex, P2-Leu and P10-Ala represent theprimary anchor residues, with P3-Asp and P6-Ile also projectingdownward toward the antigen-binding cleft. The stabilization ofthe flexible nature of this peptide by phosphorylation can beattributed to stabilizing intrapeptide interactions as well as tointeractions with Arg 65 and Lys 66 of HLA A2. In oneconformer, the phosphate group interacts with P7-His and Arg65 and it appears that the P7-His is ‘‘pulled in’’ toward thephospho-moiety compared with the nonphosphorylated com-plex. There is also a water-mediated H-bond between thephosphate group and P1-Tyr (Fig. 2 G and H).

Overall, the 6 structures have revealed a clear alteration of selfwhen phosphorylated peptides are captured and presented to Tcells—not merely via the incorporation of the phosphate moietythat alters the biophysical characteristics of the ligand but inchanges in peptide conformation and HLA A2 residues knownto influence TCR engagement (16). The phosphate moiety wasobserved in multiple conformations, suggesting that this moietycan adopt discrete conformations, thereby potentially represent-ing a ‘‘moving target’’ for TCR engagement. In each case, thephosphate sits centrally in the antigen-binding cleft as a prom-inent electronegative target for T-cell ligation (Fig. 3). Alsoshown in Fig. 3 is the putative TCR docking site on the surfaceof the pHLA complex, which indicates that the phosphate wouldbe highly accessible for interaction with CDR regions of thebound TCR.

Influence of Phosphate Group on Stability and Binding to HLA A2. Todetermine if the interactions between the phosphate group andHLA A2 residues have an impact on binding and thermostabilityof the pHLA A2 complexes, thermal denaturation curves andHLA A2 binding studies were undertaken. Good correlationbetween thermal melt curves incorporating CD measurementsof complex structure and competitive binding assays were ob-served for all 3 peptide sets (Fig. 4). The �-catenin and CDC25bcomplex stability and HLA A2 binding were essentially unaf-fected by the phosphorylation of P4-Ser and P5-Ser of therespective peptides. In contrast, the phosphorylation of P4-Serin the IRS2 peptide resulted in enhanced complex thermostability(increase of �6 °C in Tm) and improved HLA A2 binding (6.3-foldincrease in IC50). Thus, phosphorylation can increase the stabilityof the HLA A2 complex in an epitope-dependent manner.

HLA A2 Protects the Phosphate Moiety from Phosphatases. One issuethat surrounds the presence of phosphoserine within the HLAA2 binding cleft is the general labile nature of the phosphatemodification. We therefore undertook a series of experiments todetermine if the HLA A2 antigen-binding cleft afforded pro-tection to the peptide from phosphatase activity. The kinetics of

peptide dephosphorylation were assessed by mass spectrometricanalysis of phosphatase-treated samples at different time points.Consistent with the structural studies that showed the phosphateparticipated in a number of interactions with HLA A2 (TableS1), substantial protection from phosphatase activity was ob-served when the phosphopeptides were bound to HLA A2compared with the peptide in free solution (Fig. 4 C, F, and I).

ConclusionsOur structures of HLA A2 complexed to both native andphosphorylated versions of peptide epitopes provide a uniqueopportunity to visualize the impact of phosphorylation on thebound conformation of the peptide ligand and HLA A2. Inparticular, conformational adjustments to some HLA A2 resi-dues, peptide conformation, and peptide mobility were observedin an epitope-dependent manner (Fig. 1). This distinguishes ourstudy from earlier studies (15) in which only the phosphorylatedversion of the phosphopeptide epitopes was studied. Moreover,our study focuses on epitopes that are phosphorylated at the P4as well as P5 positions. Incorporation of the phosphate moietydramatically alters the electrostatic footprint of the pHLAcomplex, which, coupled to the mobility of the phosphate headgroup, is predicted to have a profound influence on T-cellrecognition.

For the IRS2 epitope, phosphorylation enhanced HLA A2binding and thermostability, although phosphorylation-dependent stabilization of the pHLA A2 was not a generaltheme. In the �-catenin epitope, phosphorylation ordered theconformation of the peptide via the introduction of electrostaticconstraints. In all cases, binding to HLA A2 shielded thephosphoserine residue from dephosphorylation by phospha-tases, suggesting that once formed, the phosphopeptide com-plexes are stable. Thus, not only are phosphopeptides trans-ported into the endoplasmic reticulum actively (10), but onassembly with HLA, they appear to be protected from phospha-tase activity, preserving the epitope for scrutiny by T cells at thecell surface. This is consistent with the ability of these peptidesto generate phosphopeptide-specific CTLs (9) and the detectionof both tumor-specific and self-phosphopeptides in the HLA A2immunopeptidome (9). This also suggests that phosphopeptideswill be present in the thymus during T-cell ontogeny, shaping theT-cell repertoire and selecting phosphopeptide-specific T cells.Thus, phosphopeptides represent a class of potential vaccinecandidates. Their presence on the surface of antigen-presentingcells not only directly reflects the altered signaling eventsoccurring in the transformed cell but generates an alteredself-pHLA landscape providing the cytotoxic T-cell effector armof the immune system an opportunity to remove malignant cellsfrom the host. Clearly, other disease states that have an impacton phosphorylation and related signaling events will also poten-tially yield distinctive pHLA landscapes that can be discerned bysurveying T lymphocytes.

Materials and MethodsPeptides. All peptides were synthesized and purified to �85% at the Bio21Peptide synthesis facility. Peptide stocks were prepared and dissolved in DMSOto a final concentration of 10–100 mg/mL.

Expression, Purification, Crystallization, and Structure Determination. Trun-cated HLA A*0201 class I heavy chain, encompassing residues 1–2745, wasexpressed as inclusion bodies using the BL21 strain of Escherichia coli asdescribed previously (17).

Crystals of all pHLA complexes were grown by the hanging drop vapordiffusion method at 20 °C using similar precipitant solutions with 2–4 mMMgCl2, 2–4 mM CdCl2, 0.1 M Hepes (pH 7.4), 100–200 mM NaCl, and 12–13%PEG3350 (vol/vol). For the IRS2 nonphospho-complex, CoCl2 was used insteadof MgCl2. Streak-seeding was required to nucleate crystals of both CDCcomplexes. Crystals appeared after 12–48 h and grew to maximal size in 7–14days. Crystals were flash-cooled to 100 K before data collection using 20%

Petersen et al. PNAS � February 24, 2009 � vol. 106 � no. 8 � 2779

IMM

UN

OLO

GY

Dow

nloa

ded

by g

uest

on

Oct

ober

7, 2

020

glycerol (vol/vol). X-ray diffraction experiments were performed using a Rik-agu RU-3HBR rotating anode generator with helium-purged OSMIC focusingmirrors coupled to an R-AXIS IV�� detector (Rikagu). All crystals belong tospace group P212121, with very similar unit cell dimensions. For a full summaryof the data collection statistics, refer to Table 1.

The structures were solved by molecular replacement using the programPhaser (18). A modified protomer of a previously solved HLA A2 structure withthe peptide residues removed was used as the search probe. Refinement wasmonitored by the Rfree value (5% of the data), using the same set of reflectionsfor all data sets. Rigid body refinement and restrained refinement wereperformed using the program Refmac (19). This was followed by simulatedannealing and individual B-factor and total least squares (TLS) refinement inPhenix (20). Model building was performed using the program Coot (21).Water molecules and peptides were built into unambiguous electron densityduring the refinement process. Cd and Co ions were modeled into strongspherical peaks (� � 8) of the 2Fo-Fc maps, and their occupancies wereadjusted manually to fit the maps. Figures were generated with Pymol(DeLano Scientific LLC) and APBS (22). For the calculation of the unbiased2Fo-Fc maps shown in Fig. 1, a single round of simulated annealing and Bfactor- and TLS refinement was performed with the peptides removed fromthe models.

CD. CD spectra were measured on a Jasco 815 spectropolarimeter using athermostatically controlled cuvette at temperatures between 30 and 90 °C.Far-UV spectra were collected and analyzed as described (17).

Competitive HLA A2 Binding Assay. Cells were treated with ice-cold citric acidfor 90 s before the binding assay to remove HLA-bound peptides. The com-petition-based peptide binding assay was performed according to van derBurg et al. (23). Briefly, 25 �L of competitor peptide (different end concen-trations) was mixed with 25 �L of fluorescence-labeled reference peptide[GILGK(FITC)VFTL, end concentration � 150 ng/mL] in a 96-well V-bottomplate. One hundred microliters of mild acid-treated JY cells (5 104/well) wasadded to the wells and incubated at 4 °C for 24 h. Cells were washed with PBScontaining 1% BSA; 10 �L of propidium iodide (1 mg/ml solution) was added,and the mean fluorescence (MF) was then measured by FACScan (BectonDickinson). Percentage inhibition of fluorescent peptide binding was calcu-lated using the following formula:

1 � MFreference�competitor peptide � MFno reference peptide� /

MFreference peptide � MFno reference peptide�� � 100%

Fig. 4. Stabilization experiments with the peptides GLLGpSPVRA, RVApSPTSGV, and YLDpSGIHSGA and their native forms. (A, D, and G) Thermal denaturationof the phosphopeptide- and nonphosphopeptide-MHC complexes using CD spectroscopy. (B, E, and H) Competition-based peptide binding assay with thephosphopeptides and nonphosphopeptides. (C, F, and I) Dephosphorylation of the single phosphopeptides and the phosphopeptide-MHC complexes by alkalinephosphatase. Nonphosphorylated peptide/complexes are shown by gray lines and phosphopeptide/complexes are shown by black lines.

2780 � www.pnas.org�cgi�doi�10.1073�pnas.0812901106 Petersen et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

7, 2

020

Phosphatase Treatment. The kinetics of peptide dephosphorylation weremeasured by MALDI-TOF mass spectrometry following alkaline phosphatasetreatment of the free phosphopeptide and phosphopeptide-HLA A2 com-plexes in solution. Five micrograms of phosphopeptide or phosphopeptide-HLA A2 complex was incubated with 5 �L of alkaline phosphatase (0.4 mg/mLfor the phosphopeptide derived from IRS2, 0.02 mg/mL for the phosphopep-tide derived from �-catenin, and 0.067 mg/mL for the phosphopeptide derivedfrom CDC25b) for 5 to 30 min. The same peptide-optimized concentrations ofphosphatase were used for the phosphopeptide-HLA A2 complexes. The

relative abundance of phosphorylated peptide was determined by MALDI-TOF mass spectrometry using an Applied Biosystems Pulsar i Q-TOF massspectrometer as described previously (24).

ACKNOWLEDGMENTS. A.W.P. is a National Health and Medical ResearchCouncil of Australia (NH&MRC) Senior Research Fellow, and J.R. is an Austra-lian Research Council Federation Fellow. This work was supported byNH&MRC Project Grants 491117 and 508927 and National Institutes of HealthGrant GM057428-06.

1. Johnson DA (2008) Synthetic TLR4-active glycolipids as vaccine adjuvants and stand-alone immunotherapeutics. Curr Top Med Chem 8:64–79.

2. Dorn A, Kippenberger S (2008) Clinical application of CpG-, non-CpG-, and antisenseoligodeoxynucleotides as immunomodulators. Curr Opin Mol Ther 10:10–20.

3. Morita CT, Jin C, Sarikonda G, Wang H (2007) Nonpeptide antigens, presentationmechanisms, and immunological memory of human Vgamma2Vdelta2 T cells: Discrim-inating friend from foe through the recognition of prenyl pyrophosphate antigens.Immunol Rev 215:59–76.

4. Betser-Cohen G, et al. (2006) Reduced KIR2DL1 recognition of MHC class I moleculespresenting phosphorylated peptides. J Immunol 176:6762–6769.

5. Mandell JW (2003) Phosphorylation state-specific antibodies: Applications in investi-gative and diagnostic pathology. Am J Pathol 163:1687–1698.

6. Datta SK (2003) Major peptide autoepitopes for nucleosome-centered T and B cellinteraction in human and murine lupus. Ann N Y Acad Sci 987:79–90.

7. Utz PJ, Hottelet M, van Venrooij WJ, Anderson P (1998) Association of phosphorylatedserine/arginine (SR) splicing factors with the U1-small ribonucleoprotein (snRNP) au-toantigen complex accompanies apoptotic cell death. J Exp Med 187:547–560.

8. Zarling AL, et al. (2000) Phosphorylated peptides are naturally processed and pre-sented by major histocompatibility complex class I molecules in vivo. J Exp Med192:1755–1762.

9. Zarling AL, et al. (2006) Identification of class I MHC-associated phosphopeptides astargets for cancer immunotherapy. Proc Natl Acad Sci USA 103:14889–14894.

10. Andersen MH, et al. (1999) Phosphorylated peptides can be transported by TAPmolecules, presented by class I MHC molecules, and recognized by phosphopeptide-specific CTL. J Immunol 163:3812–3818.

11. Scanlan MJ, Gure AO, Jungbluth AA, Old LJ, Chen YT (2002) Cancer/testis antigens: Anexpanding family of targets for cancer immunotherapy. Immunol Rev 188:22–32.

12. Mayordomo JI, et al. (1996) Therapy of murine tumors with p53 wild-type and mutantsequence peptide-based vaccines. J Exp Med 183:1357–1365.

13. Williamson NA, Rossjohn J, Purcell AW (2006) Tumors reveal their secrets to cytotoxicT cells. Proc Natl Acad Sci USA 103:14649–14650.

14. Gottschalk S, Heslop HE, Rooney CM (2005) Adoptive immunotherapy for EBV-associated malignancies. Leuk Lymphoma 46:1–10.

15. Mohammed F, et al. (2008) Phosphorylation-dependent interaction between antigenicpeptides and MHC class I: A molecular basis for the presentation of transformed self.Nat Immunol 9:1236–1243.

16. Tynan FE, et al. (2005) T cell receptor recognition of a ‘super-bulged’ major histocom-patibility complex class I-bound peptide. Nat Immunol 6:1114–1122.

17. Webb AI, et al. (2004) Functional and structural characteristics of NY-ESO-1-relatedHLA A2-restricted epitopes and the design of a novel immunogenic analogue. J BiolChem 279:23438–23446.

18. Storoni LC, McCoy AJ, Read RJ (2004) Likelihood-enhanced fast rotation functions. ActaCrystallogr D 60(Pt 3):432–438.

19. Murshudov GN, Vagin AA, Dodson EJ (1997) Refinement of macromolecular structuresby the maximum-likelihood method. Acta Crystallogr D 53(Pt 3):240–255.

20. Adams PD, et al. (2002) PHENIX: Building new software for automated crystallographicstructure determination. Acta Crystallogr D 58:1948–1954.

21. Emsley P, Cowtan K (2004) Coot: Model-building tools for molecular graphics. ActaCrystallogr D 60(Pt 1):2126–2132.

22. Baker NA, Sept D, Joseph S, Holst MJ, McCammon JA (2001) Electrostatics of nanosystems:Application to microtubules and the ribosome. Proc Natl Acad Sci USA 98:10037–10041.

23. van der Burg SH, et al. (1995) An HLA class I peptide-binding assay based on competitionfor binding to class I molecules on intact human B cells: Identification of conserved HIV-1polymerase peptides binding to HLA-A *0301. Human Immunol 44:189–198.

24. Purcell AW, et al. (2001) Quantitative and qualitative influences of tapasin on the classI peptide repertoire. J Immunol 166:1016–1027.

25. Garboczi DN, et al. (1996) Structure of the complex between human T-cell receptor,viral peptide and HLA-A2. Nature 384:134–141.

Petersen et al. PNAS � February 24, 2009 � vol. 106 � no. 8 � 2781

IMM

UN

OLO

GY

Dow

nloa

ded

by g

uest

on

Oct

ober

7, 2

020