Vaccination with NY-ESO-1 protein and CpG inMontanide induces integrated antibody/Th1 responsesand CD8 T cells through cross-primingDanila Valmori*†, Naira E. Souleimanian*, Valeria Tosello*, Nina Bhardwaj‡, Sylvia Adams‡, David O’Neill‡,Anna Pavlick‡, Juliet B. Escalon‡, Crystal M. Cruz‡, Angelica Angiulli‡, Francesca Angiulli‡, Gregory Mears§,Susan M. Vogel§, Linda Pan¶, Achim A. Jungbluth¶, Eric W. Hoffmann¶, Ralph Venhaus¶, Gerd Ritter¶,Lloyd J. Old¶�, and Maha Ayyoub*†

*Ludwig Institute Clinical Trial Center, Columbia University, New York, NY 10032; ‡New York University School of Medicine, New York, NY 10016; §Divisionof Medical Oncology, Columbia University Medical Center, New York, NY 10032; and ¶Ludwig Institute for Cancer Research, New York, NY 10158

Contributed by Lloyd J. Old, April 12, 2007 (sent for review February 22, 2007)

The use of recombinant tumor antigen proteins is a realisticapproach for the development of generic cancer vaccines, but thepotential of this type of vaccines to induce specific CD8� T cellresponses, through in vivo cross-priming, has remained unclear. Inthis article, we report that repeated vaccination of cancer patientswith recombinant NY-ESO-1 protein, Montanide ISA-51, and CpGODN 7909, a potent stimulator of B cells and T helper type 1(Th1)-type immunity, resulted in the early induction of specificintegrated CD4� Th cells and antibody responses in most vacci-nated patients, followed by the development of later CD8� T cellresponses in a fraction of them. The correlation between antibodyand T cell responses, together with the ability of vaccine-inducedantibodies to promote in vitro cross-presentation of NY-ESO-1 bydendritic cells to vaccine-induced CD8� T cells, indicated thatelicitation of NY-ESO-1-specific CD8� T cell responses by cross-priming in vivo was associated with the induction of adequatelevels of specific antibodies. Together, our data provide clearevidence of in vivo cross-priming of specific cytotoxic T lympho-cytes by a recombinant tumor antigen vaccine, underline theimportance of specific antibody induction for the cross-priming tooccur, and support the use of this type of formulation for thefurther development of efficient cancer vaccines.

cancer vaccine

A key step in the development of generic cancer vaccines isthe implementation of vaccination strategies allowing for

the consistent induction of immune responses to tumor antigens.In this respect, the choice of appropriate antigens, based on boththe frequency and the specificity of their expression in cancertissues, is of paramount importance. The group of cancer/testisantigens (CTA) (1, 2), including the NY-ESO-1 antigen (3), isemerging among the most promising candidates. Because manyCTA are not expressed on the surface of cancer cells but ratherintracellularly, it is important that vaccination induces specificCD8� T cells able to directly recognize antigen-expressing tumorcells. Recombinant proteins can be produced in large scale andat relatively low cost, are commonly used in the development ofantiviral vaccines, and are therefore attractive candidate anti-tumor vaccines. The potential of tumor antigen recombinantprotein vaccines, however, relies on their ability not only to elicitantibody (Ab) and CD4� T cell responses but also to efficientlyprime naive CD8� T cells through cross-priming (4), whichgenerally is inefficient during spontaneous immune responses totumor antigens (5). Professional antigen-presenting cells (APCs)detect pathogens through a variety of receptors such as theToll-like receptors (TLR), which recognize pathogen-associatedmolecular patterns including CpG dinucleotides within definedflanking sequences (CpG ODN) (6). Synthetic CpG ODN ableto trigger TLR9 are potent vaccine adjuvants, stimulating T

helper type 1 (Th1)-type immunity (7). In humans, they candirectly activate B lymphocytes and plasmacytoid dendritic cellsand also indirectly activate myeloid dendritic cells (mDCs),increasing antigen cross-presentation and stimulating adaptiveimmune responses (8–10).

In this study, we have assessed the immune response elicitedby repeated vaccination with a NY-ESO-1 recombinant protein(rNY-ESO-1) administered with CpG 7909 in a water–oil emul-sion with Montanide ISA-51. We show that cancer patientsreceiving this vaccine developed integrated Ab and CD4� T cellresponses to NY-ESO-1 at an early phase of the vaccinationprotocol. A fraction of the patients also developed specific CD8�

T cell responses at a later time point. Assessment of thecorrelation between the development of Ab and T cell responsessuggested that the presence of sufficient levels of NY-ESO-1-specific antibodies was determinant for the cross-priming ofCD8� T cells to occur in vivo. In line with this concept, we foundthat in vitro cross-presentation of NY-ESO-1 protein to vaccine-induced CD8� T cells by dendritic cells was enhanced byvaccine-induced Ab.

ResultsSerological Response to Vaccination with rNY-ESO-1, CpG 7909, andMontanide ISA-51. Study patients received four s.c. injections ofNY-ESO-1/Montanide/CpG vaccine at 3-week intervals. Thestudy included two arms receiving 100 or 400 �g of rNY-ESO-1per injection, together with 2.5 mg of CpG emulsified in Mon-tanide. CpG 7909 belongs to the CpG-B class, which potentlystimulates B cells (11). All patients developed significant sero-logical responses to NY-ESO-1 (Fig. 1A). Specific IgG responsesbecame significant after the second injection and further in-creased after the third and fourth injections, without generallyreaching a plateau. Specific IgG titers were variable amongpatients: at the last time point, NY-ESO-1 reciprocal Ab titershad reached an average 42.429 � 29.080 in the 400 �g group and24.091 � 10.737 in the 100 �g group. Serological responses to an

Author contributions: D.V., L.J.O., and M.A. designed research; D.V., N.E.S., V.T., N.B., S.A.,D.O., A.P., J.B.E., C.M.C., A.A., F.A., G.M., S.M.V., A.A.J., and M.A. performed research; L.P.,E.W.H., R.V., G.R., and L.J.O. contributed new reagents/analytic tools; D.V. and M.A.analyzed data; and D.V. and M.A. wrote the paper.

The authors declare no conflict of interest.

Abbreviations: CTL, cytotoxic T lymphocyte; APC, antigen-presenting cell; TLR, Toll-likereceptors; Th, T helper; mDC, myeloid dendritic cell; PBMC, peripheral blood mononuclearcell; ADCC, antibody-dependent cellular cytotoxicity.

†To whom correspondence may be sent at the present address: Institut National de laSante et de la Recherche Medicale, Unite 601, Centre de Lutte Contre le Cancer ReneGauducheau, Boulevard Jacques Monod, 44800 Saint Herblain, France. E-mail:danila.valmori@univ-nantes.fr or maha.ayyoub@univ-nantes.fr.

�To whom correspondence may be addressed. E-mail: lold@licr.org.

© 2007 by The National Academy of Sciences of the USA

www.pnas.org�cgi�doi�10.1073�pnas.0703395104 PNAS � May 22, 2007 � vol. 104 � no. 21 � 8947–8952

IMM

UN

OLO

GY

Dow

nloa

ded

by g

uest

on

Janu

ary

9, 2

020

unrelated recombinant protein (MAGE-A4) generally were lowin both groups (4.957 � 3.594 in the group receiving 400 �g and2.735 � 2.519 in the group receiving 100 �g).

Isotype of Vaccine-Induced IgG and Mapping of B Cell Epitopes. Toinvestigate further the serological responses induced by thevaccine, we assessed the isotype of vaccine-induced IgG by usingisotype-specific antibodies. As illustrated in Fig. 1B, the vaccinepreferentially induced specific IgG of types 1 and 3, which areassociated, in humans, with type 1 immune responses. Wemapped the linear B cell epitopes recognized by the vaccine-induced Ab with a panel of peptides spanning the proteinsequence (12). The analysis identified a dominant region locatedin the central part of the protein and comprising amino acids 80–110, which was recognized strongly by Abs from all patients (Fig.1C). Consistent with the previous identification of B cell epitopeslocated in the N-terminal half of the protein and recognized byspecific Abs from cancer patients with spontaneous serologicalresponses to NY-ESO-1 (12, 13), the immune sera from vacci-nated patients reacted to peptides in the amino acids 1–70 regionof the protein. A lower level of reactivity was detected towardsequences located in the C-terminal half of the protein, betweenamino acids 160 and 180.

Assessment of T Cell Responses After in Vitro Stimulation. Weinitially assessed the induction of specific T cells after in vitrostimulation of pre- and postimmune samples with a pool ofoverlapping peptides spanning the protein sequence, followed byquantification of specific IFN-�-producing cells by intracellularstaining (14, 15) (Fig. 2A). The response was considered signif-icant if the frequency of T cells detected in at least onepostvaccine sample exceeded by 3-fold that found in the baselinesample. CD4� T cell responses were detectable in 17 of 18patients and developed early, in general becoming detectablealready at week 4 (Fig. 2B). CD8� T cell responses weredetectable in 9 of 18 patients (4/7 in the 400 �g group and 5/11in the 100 �g group) and generally developed later. We usedstimulated cultures to map the protein regions recognized byCD4� and CD8� T cells, by using single peptides in theNY-ESO-1 pool (Fig. 2C). Similar to what was found previouslyin patients with spontaneous responses (14–16), the majority of

CD4� T cells recognized sequences located in two distinctregions of the protein, corresponding to peptides 81–100 and119–143. We found minor reactivity toward some other peptides,including 151–170. The sequences recognized by CD8� T cellsmostly were located between amino acids 81 and 110, as indi-cated by the reactivity of the majority of these cells to theoverlapping peptides 81–100 and 91–110. The relative reactivityto these two peptides varied among patients, indicating thepresence of multiple epitopes in this region, consistent withprevious results in patients with spontaneous responses as wellas in patients immunized with other NY-ESO-1 full-lengthvaccines (15, 17, 18).

Ex Vivo Assessment of T Cell Responses. Next, total pre- andpostimmune peripheral blood mononuclear cell (PBMC) sam-ples were stimulated directly ex vivo with the NY-ESO-1 peptidepool and assessed for the presence of specific T cells by usingintracellular cytokine staining together with cell-surface stainingto assess phenotype. NY-ESO-1-specific CD4� and CD8� T cellswere detectable ex vivo after vaccination but not before vacci-nation (Fig. 3A) We detected significantly increased levels ofNY-ESO-1-specific T cells in postvaccination samples from 12 ofthe 18 patients in the case of CD4� T cells and in 7 patients in

Fig. 1. Serological responses. (A) Serological responses were assessed byELISA at baseline and at the indicated study week after vaccination. (B) Theisotype of vaccine-induced IgG (week 12, serum dilution of 1:100) was assessedby ELISA using isotype-specific antibodies. (C) Linear B cell epitopes weredetermined by using patients’ immune sera and a panel of 30-aa-long pep-tides spanning the protein sequence.

Fig. 2. Assessment of CD4� and CD8� T cell responses after in vitro stimu-lation. (A) The presence of specific CD4� and CD8� T cells in stimulated culturesfrom vaccinated patients was assessed by intracellular staining with anti-IFN-�monoclonal Ab after stimulation in the absence or presence of the NY-ESO-1peptide pool. Numbers in the upper right quadrant are the percentages ofIFN-�-producing cells in CD4� or CD8� T cells. As an example, data are shownfor one responder patient. (B) Summary of the results obtained for all patients.Values correspond to the percentage of CD4� or CD8� T cells specificallyproducing IFN-� in response to stimulation with the peptide pool. (C) Cultureswere stimulated with single peptides in the NY-ESO-1 pool. The proportion ofIFN-�-producing cells was assessed by intracellular staining as above. Symbolsare as given in the Fig. 1 key.

8948 � www.pnas.org�cgi�doi�10.1073�pnas.0703395104 Valmori et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

9, 2

020

the case of CD8� T cells (Fig. 3B). It is noteworthy that 5 of 6patients in whom CD4� T cell responses were not detectable exvivo were among nonresponders or low responders after in vitrostimulation. In addition, 6 of 7 patients with CD8� T cellresponses detectable ex vivo were among the 9 patients in whomCD8� T cell responses were detectable after in vitro stimulation.Vaccine-induced CD4� T cells mostly were composed of effec-tor-memory (CD45RA� CCR7�) and, to a lower extent, central-memory (CD45RA� CCR7�) T cells (19) (Fig. 3C). A large partof these cells secreted IFN-� only, whereas lower proportionseither cosecreted IFN-� and IL-2 or secreted IL-2 only. Vaccine-induced CD8� T cells also were composed mostly of effector-memory T cells, did not contain many central-memory T cells,and contained instead a significant proportion of terminallydifferentiated effectors (CD45RA�CCR7�). The cytokine pro-file of vaccine-induced CD4� and CD8� T cells was clearly ofTh1 type, because we failed to detect any significant levels ofIL-10-, IL-4-, or IL-17-secreting cells (data not shown).

Recognition of NY-ESO-1 by Vaccine-Induced T Cells and Ab-AidedCross-Presentation. To address the ability of vaccine-inducedCD4� and CD8� T cells to recognize exogenously and endog-

enously processed NY-ESO-1 antigen, we isolated them frompeptide-stimulated cultures by IFN-�-guided cell sorting (Fig.4A). We used the isolated populations to assess the recognitionof autologous mDCs incubated with the rNY-ESO-1 or electro-porated with a plasmid encoding NY-ESO-1 (Fig. 4B). TherNY-ESO-1 was recognized efficiently by CD4� T cells butpoorly by CD8� T cells. In contrast, CD8� T cells but not CD4�

T cells recognized the endogenous antigen. These results suggestthat, in this setting, autologous professional APCs are likely toinitiate the immune response by priming CD4� Th cells afterprocessing and presentation of the rNY-ESO-1 protein. Toaddress the ability of CD8� T cells to recognize tumor cells, we

Fig. 3. Ex vivo assessment of CD4� and CD8� T cell responses. (A) Dot plotsgated on CD3� cells are shown for a high-responder patient. Numbers in theupper right quadrant are the percentages of IFN-�-secreting cells among CD4�

T cells (Upper) and CD8� T cells (Lower). (B) Summary of the results obtainedfor all patients. Symbols are as given in the Fig. 1 key. For each sample, valuesobtained in the absence of the NY-ESO-1 peptide pool were subtracted.Dashed lines represent the mean values of baseline samples plus three timestheir standard deviation. The baseline sample from patient N4, who haddetectable NY-ESO-1-specific CD8� T cells before vaccination as shown in Fig.2 and Table 1, was excluded from this calculation. Frequencies of IFN-�-secreting cells above this value were considered significant. (C) Phenotype andcytokine production of vaccine-induced CD4� and CD8� T cells.

Fig. 4. Recognition of NY-ESO-1 by vaccine-induced T cells and Ab-aidedcross-presentation. (A) Vaccine-induced CD4� and CD8� T cell populations en-riched by cytokine-secretion-guided sorting were assessed by intracellular IFN-�staining after stimulation in the absence or presence of the NY-ESO-1 peptidepool. (B) Recognition of autologous mDCs by CD4� and CD8� T cells was assessedafter incubation with serial dilutions of rNY-ESO-1 or after electroporation witha plasmid encoding NY-ESO-1. (C) The MHC class I allele restricting antigenrecognition by CD8� T cells was determined upon transfection of COS-7 cells withplasmids encoding the alleles of the patient. Recognition of endogenous NY-ESO-1 was assessed upon cotransfection of COS-7 cells with plasmids encodingNY-ESO-1 and HLA-B35 or by transfecting the tumor cell lines NA8-MEL (NY-ESO-1�) and HT1080 (NY-ESO-1�) with a plasmid encoding HLA-B35. (D) Cross-presentation, by autologous mDCs, of rNY-ESO-1 (10 �g/ml) or immune com-plexes formed either with a murine NY-ESO-1-specific monoclonal Ab (ES121, 10�g/ml) or with pre- and postimmune serum at the indicated dilution to vaccine-induced CD8� T cells. In B and C, antigen recognition was assessed by measure-ment of IFN-� secretion in the culture supernatant.

Valmori et al. PNAS � May 22, 2007 � vol. 104 � no. 21 � 8949

IMM

UN

OLO

GY

Dow

nloa

ded

by g

uest

on

Janu

ary

9, 2

020

identified the MHC class I allele restricting antigen recognitionof CD8� T cells from one responder patient and assessed itsability to recognize antigen-expressing tumor cells transfectedwith the appropriate allele. Transfection of COS-7 cells with anHLA-B35-encoding plasmid, but not with plasmids encodingother patient’s HLA alleles, resulted in peptide presentation toCD8� T cells (Fig. 4C). In addition, both COS-7 cells cotrans-fected with NY-ESO-1 and HLA-B35 and NY-ESO-1-expressing tumor cells transfected with HLA-B35 were recog-nized specifically by CD8� T cells.

Enhancement of Antigen Cross-Presentation by Vaccine-Induced Ab.The data above provided evidence that vaccination resulted inthe consistent induction of CD4� T cell and Ab responses in mostpatients, as well as of specific CD8� T cells in a fraction of them(summarized in Table 1). The in vivo kinetic of the responsesclearly showed that CD4� T cells and Abs developed early,whereas CD8� T cells generally developed later. In addition, thedevelopment of CD8� T cell responses generally correlated withmedium to high titers of NY-ESO-1 Ab. This, together with therelatively poor recognition of the rNY-ESO-1 alone after cross-presentation by dendritic cells, suggested an involvement ofNY-ESO-1 Ab in the in vivo cross-priming of CD8� T cells,through formation of immune complexes. To address this point,mDCs from a responder patient were pulsed with the NY-ESO-1recombinant protein for 2 h and then assessed for antigenrecognition by autologous CD8� T cells. Under these conditions,no significant cross-presentation was detected (Fig. 4D). How-ever, dendritic cell incubation with immune complexes gener-ated by mixing NY-ESO-1 recombinant protein with the NY-ESO-1-specific monoclonal Ab ES121 resulted in efficientantigen cross-presentation. In addition, preincubation of NY-ESO-1 protein with postvaccine serum but not with prevaccineserum, at various dilutions, also resulted in efficient antigencross-presentation. Together, these results are in line with theconcept that the induction of NY-ESO-1-specific Ab by the

vaccine formulation used in this study may be a determiningfactor for the induction of CD8� T cells, through Ab-aidedcross-priming in vivo.

DiscussionNY-ESO-1 is a nonmutated self-antigen frequently expressed inhuman tumors that often induces spontaneous immune responsesin patients bearing antigen-expressing tumors. We have selectedNY-ESO-1 as a prototype tumor antigen for the development ofgeneric cancer vaccines and, under the sponsorship of the CancerVaccine Collaborative (www.cancerresearch.org), are assessing theimmunogenicity of NY-ESO-1-based candidate vaccines in early-phase clinical trials. Among the immunogens tested is the rNY-ESO-1 protein, in various formulations (17, 18). In this study, wehave obtained direct evidence that immunization with rNY-ESO-1administered with Montanide and CpG induces integrated Ab andCD4� T cell responses in the majority of vaccinated patients andcross-primes CD8� T cells in a significant fraction of them. Al-though NY-ESO-1 expression in tumor specimens could not beassessed for all patients, we failed to detect any correlation betweenantigen expression and responsiveness to vaccination (Table 1). Itis noteworthy that, with the exception of one patient, no significantserological reactivity to NY-ESO-1 was detectable before vaccina-tion, indicating the ability of the vaccine to prime specific naıve Bcells. This finding is in line with the recent proposal by Ruprecht etal. that triggering of TLR9 by CpG on human naıve B cells acts asa third signal, required, in addition to B cell receptor stimulationand CD4� T cell help, for their full differentiation into IgG-secreting cells (20). Montanide may contribute significantly to theimmunogenicity of the formulation, through its known effects, suchas depot effect, slow antigen release, and recruitment of APCs atthe injection site. In support of this data, in a separate study,administration of NY-ESO-1 protein and CpG without Montanidewas less immunogenic (E. Jager, personal communication). Ad-ministration of CpG in combination with Montanide, however, islikely to be responsible for the high Ab titers detected in this study

Table 1. Patient characteristics, antigen dose, summary of vaccine-related toxicity, and vaccine-induced immune responses

Patient Disease* StageBefore study

status†

Antigendose, �g

Toxicity‡

(grade)

Ab§ CD4¶ CD8

Before After Before After Before After

N14 Melanoma IIB NED 100 L(1), S(1) � � � � � �

N11 Melanoma IIIB NED 100 L(1), S(1) � � � � � �

C3 Breast cancer I NED 100 L(2) � � � � � �

C4 Breast cancer IIA NED 100 L(1), S(1) � � � �� � �

N10 Melanoma IVA NED 400 L(1) � � � � � �

C7 Melanoma III NED 100 L(1), S(1) � �� � � � �

N13 Melanoma IIIA NED 400 L(1), S(1) � �� � � � �

N5 Melanoma IVB NED 400 L(1) � �� � � � �

C1 Sarcoma II NED 400 L(2) � �� � �� � �

N4 Melanoma IVA NED 100 L(1), S(1) � � � � � ��

C5 Sarcoma III M (lung) 100 L(2), S(2) � �� � �� � �

N2 Melanoma IIIB NED 100 L(1), S(1) � �� � �� � �

N7 Melanoma IIIB NED 100 L(1), S(1) � �� � �� � �

N9 Ovarian cancer IIIC NED 400 L(1), S(1) � �� � �� � ��

N3 Melanoma IIIB NED 100 L(1), S(1) � �� � ��� � ��

C6 Sarcoma IIIB NED 400 L(1), S(1) � ��� � �� � ��

N8 Melanoma IIC NED 100 L(1), S(1) � ��� � �� � ��

C2 Breast cancer IIA NED 400 L(2), S(2) � ��� � �� � ���

*NY-ESO-1 expression in the tumor was not an entrance criterion for this study. However, 10 of 18 patients were tested for NY-ESO-1 expression byimmunohistochemistry. Patients N4 and N9 were NY-ESO-1�. Patients C3, C4, C5, C6, N2, N3, N7, and N13 were NY-ESO-1�.

†NED, no evidence of disease; M, metastatic disease.‡Adverse events with possible, probable, or definite relationship to study drug. L, local toxicity at injection site, included redness, discomfort, itching, induration,erythema, and pain. S, systemic toxicity, included fever, chills, fatigue, cold-like symptoms, sweats, muscle aches, and joint pain.

§Ab (week 12 reciprocal serum titer): �, �100; �, �100 to 20,000; ��, �20,000 to 40,000; ���, �40,000.¶CD4 and CD8 (percentage of IFN-�� cells after in vitro stimulation): �, �0.3%; �, �0.3% to 1%; ��, �1% to 10%; ���, �10%.

8950 � www.pnas.org�cgi�doi�10.1073�pnas.0703395104 Valmori et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

9, 2

020

and to the skewing of the induced immunological response towarda Th1 type, attributable to the induction of proinflammatorycytokines and type 1 IFNs, in contrast to incomplete Freund’sadjuvant alone (equivalent to Montanide), which induced a Th2-type bias (11, 21). Consistent with this concept, vaccine-inducedIgG predominantly were IgG1 and IgG3, the most prominent IgGtype 1 human subclasses, involved in functions such as complementactivation, mediation of antibody-dependent cellular cytotoxicity(ADCC), and opsonization (22).

After one round of in vitro stimulation, vaccine-induced CD4� Tcell responses were detectable in most vaccinated patients, andCD8� T cell responses were detectable in a fraction of them. Inmost cases, their reactivity was restricted to one or few peptides inthe protein, corresponding to immunodominant regions identifiedin our previous studies (14, 15). The immunodominant regionrecognized by cross-primed CD8� T cells (between amino acids 81and 110) was one of the two main regions also recognized by CD4�

T cells and the most efficiently recognized by Ab. We recently havereported the identification of overlapping clusters of MHC class Iand class II binding sequences in immunodominant regions ofNY-ESO-1 that contain a high density of proteasomal cleavage sites(14, 15). This interesting phenomenon provides the molecular basesfor the development of integrated Ab, Th, and cytotoxic T lym-phocyte (CTL) responses around short immunodominant regions.Based on these findings, we have proposed that CTL epitopes in theNY-ESO-1 immunodominant 81–110 region may be generatedefficiently through cross-priming, a hypothesis now supported bythe findings of the present study. In addition, the overlappingbetween regions concentrating B and T cell immunodominantepitopes, suggests a role for Ab-mediated epitope protection duringantigen processing.

Both vaccine-induced CD4� and CD8� T cell populations ex-hibited ex vivo a clear Th1 profile and contained polyfunctionalpopulations producing IL-2 and/or IFN-�. They mostly exhibited aneffector-memory phenotype, although CD4� T cell populationscontained detectable proportions of central memory cells andCD8� T cell populations of so-called terminally differentiatedeffectors. The ex vivo frequencies of vaccine-induced NY-ESO-1-specific T cells were variable among patients, and, with the excep-tion of one high-responder patient, they remained relatively low inmost patients. The ex vivo frequency of CD4� T cells detected in ourstudy is in the range of those recently reported after vaccination ofnonhuman primates with HIV Gag protein, Montanide, and CpGODN (C class) (23). Interestingly, in the study from Wille-Reece etal. (23), TLR7/8 agonists stimulated higher responses than CpGODN. In addition, in their study, further immunization withGag-encoding recombinant adenovirus resulted in a high increasein the frequency of CD8� T cells, suggesting that prime/boostregimens alternating rNY-ESO-1 and NY-ESO-1-encoding recom-binant viruses (currently in development as part of the CancerVaccine Collaborative program) (18) could lead to an optimalexpansion of specific CD8� T cells.

The role of CpG in the induction of primary and memory T cellresponses has not been elucidated completely. Based on in vitroexperiments, Rothenfusser et al. (24) have proposed that class ACpG could selectively enhance memory CD8� T cell responses andcytotoxicity, whereas class B CpG could be superior for primingCD8� T cells. In support of this data, in a recent study, Speiser etal. (25) reported evidence of induction of CD8� T cell responses inmelanoma patients, including some with no detectable preimmuneresponses, by vaccination with a Melan-A peptide analog, Mon-tanide, and CpG. In our study, priming of CD4� T cells occurredat an early phase of the vaccination protocol, in contrast to CD8�

T cell responses that became detectable only at a later stage. Thisdelayed appearance, which occurred concomitant with the titer ofNY-ESO-1-specific Abs reaching high levels in responder patients,suggested a role for vaccine-induced Abs in promoting CD8� T cellcross-priming through Ab-aided cross-presentation (26–29). In

support of this, postimmune sera from vaccinated patients wereable to promote efficiently cross-presentation of NY-ESO-1 tovaccine-induced CD8� T cells in vitro. In addition to their titers, theinduction of type 1 IgG Ab is likely of importance. Fc�R is a familyof glycosylated membrane proteins that includes members thatdiffer largely with respect to their capacity to bind IgG (30, 31). Forexample, Fc�RI (CD64) binds monomeric IgG with high affinity,but it signals only if the latter is cross-linked by specific polymericligands. In contrast, Fc�RII (CD32) binds monomeric IgG poorlybut binds immune complexes with very high affinity. IgG3 and IgG1preferentially interact with Fc�RIIa, which is a potent leukocyteactivator and can stimulate the release of high levels of inflamma-tory cytokines. Thus, the Ab isotypes induced by our vaccine, inparticular IgG3, may be effective at inducing dendritic cell activa-tion and maturation, rapid internalization, concentration, andtransport of the antigen to the appropriate endosomal/lysosomalcompartments, ultimately enhancing the induction of T cells.

As in the case of NY-ESO-1, specific Ab responses to tumor-associated antigens often develop during spontaneous immuneresponses to tumors. The biological role of these Abs in thedevelopment of the disease, however, is poorly understood. Themechanism of action in vivo of some antitumor Abs currently usedfor the treatment of breast cancer (anti-Her2/neu) or B celllymphomas (anti-CD20) remains unclear but generally is consid-ered to be through ADCC (32). Possibly complementary to ADCC,Ab-aided cross-priming of CTLs specific for tumor antigens alsocould contribute to the antitumor effect of these therapies. In amurine model of melanoma, passive transfer of Abs against themelanoma antigen gp75 has been shown to be highly effective ininhibiting tumor growth (33) in a Fc�R-dependent manner (34). Ina different mouse model, using EL-4 lymphoma cells transfectedwith the model antigen OVA, Dyall et al. (35) have provided furtherevidence for Fc�R-mediated Ab-dependent tumor eradication andshown that CD8� T cells are the critical effector population in thissystem. Additional evidence for the role of antitumor Abs inpromoting cross-priming of tumor-specific CTLs has been providedby an in vitro study with human cells in which Dhodapkar andcolleagues (36) demonstrated that coating myeloma cells withanti-syndecan-1 Abs strongly enhanced cross-presentation andcross-priming of CTLs specific for the NY-ESO-1 and MAGE-A3antigens. Adding to these previous findings, our data indicate thatAb-aided cross-priming of NY-ESO-1-specific CTLs, and possiblyof other human tumor antigens, could be achieved through activeimmunization with recombinant proteins.

Along with Abs, other factors are likely to contribute, bothdirectly and indirectly, to promote in vivo cross-priming aftervaccination with the formulation used in our study. Namely,although class B CpG may not be optimal for inducing the pro-duction of IFN-�, the latter still could play a significant role by bothstimulation of B cell activating factor (BAFF) (37) and activation ofmDCs (38). Also, vaccine-induced CD4� T cells could play a rolein the activation of mDCs through CD40–CD40L interactions (28).In conclusion, the vaccine formulation used in this study seems tobe highly immunogenic and able to stimulate integrated Ab and Tcell responses, including tumor-reactive CD8� T cells, by cross-priming in an Ab-aided fashion. On the other hand, the relativelydelayed development of specific CD8� T cell responses requiringrepeated vaccine injections, along with their detection in only afraction of vaccinated patients, suggests that additional injections,a combination of Montanide/CpG with other TLR ligands, or theuse of prime/boost strategies alternating different NY-ESO-1-based immunogens may be required to optimize their induction.Based on our data, administration of NY-ESO-1-specific Absbefore or concomitant with vaccination or administration of pre-formed activating Ab–protein immune complexes in combinationwith adjuvant also may be promising approaches for the develop-ment of cancer vaccines of increasing efficacy.

Valmori et al. PNAS � May 22, 2007 � vol. 104 � no. 21 � 8951

IMM

UN

OLO

GY

Dow

nloa

ded

by g

uest

on

Janu

ary

9, 2

020

Materials and MethodsClinical Study and Patient Population. The study protocol wasapproved by the Institutional Review Boards of ColumbiaUniversity and New York University Medical Centers. Patientsamples were obtained by means of written informed consent.The patients enrolled had histological diagnoses of cancer typesknown to express NY-ESO-1. However, demonstration of ex-pression of NY-ESO-1 in the patient’s tumor specimen was notan inclusion criterion. The primary end point of the study was todetermine the safety and immunogenicity of the vaccine. Tox-icity and adverse events were graded according to the NationalCancer Institute’s Common Terminology Criteria for AdverseEvents scale (version 3.0; accessed December 12, 2003).

Assessment of Serological Responses. Serological responses wereassessed by ELISA using NY-ESO-1 or MAGE-A4 recombinantproteins produced in Escherichia coli (12, 39). Sera were assessedover a range of dilutions from 1/100 to 1/100,000. Titers werecalculated as the serum dilution giving 50% of maximal opticaldensity obtained by using a standard positive serum. IgG isotypeswere determined by ELISA using the same protocol and antibodiesspecific for IgG1 (BD Biosciences, San Jose, CA) and IgG2, IgG3,and IgG4 (Southern Biotech, Birmingham, AL).

Assessment of T Cell Responses. CD4� and CD8� cells were enrichedfrom PBMCs by magnetic cell sorting using miniMACS (MiltenyiBiotec Inc., Bergisch Gladbach, Germany) and stimulated withirradiated autologous APCs in the presence of the NY-ESO-1peptide pool (each 2 �M; NeoMPS Inc., San Diego, CA), rhIL-2 (10IU/ml), and rhIL-7 (10 ng/ml). At day 8, cultures were tested forintracellular IFN-� secretion after stimulation, during 4 h, and inthe absence or presence of the peptide pool or of individualpeptides (14, 15). For ex vivo assessments, cryopreserved totalPBMCs were thawed, rested overnight, and stimulated for 7 h in the

absence or presence of the NY-ESO-1 peptide pool. Brefeldin Awas added 2 h after the beginning of the incubation. At the end ofthe incubation, cells were stained with antibodies directed againstsurface markers (CD3, CD4, CD8, CD45RA, and CCR7), fixed,permeabilized, and stained with cytokine-specific antibodies.

Assessment of Antigen Recognition by Specific T Cells. NY-ESO-1-specific T cells were isolated from peptide-stimulated cultures bycytokine-guided magnetic cell sorting (Miltenyi Biotec Inc.), stim-ulated as polyclonal cultures in the presence of phytohemagglutinin,allogeneic irradiated PBMCs, and rhIL-2 (100 units/ml), and usedin antigen recognition assays. mDCs were obtained by culture ofisolated CD14� cells during 5 days in the presence of rhIL-4 andrhGM-CSF. mDCs were either incubated overnight in the presenceof the NY-ESO-1 protein and washed before use in recognitionassays or incubated for 2 h in the presence of the rNY-ESO-1 orNY-ESO-1 immune complexes, washed, cultured overnight, andused in recognition assays. Where indicated, mDCs were trans-fected with a NY-ESO-1-expressing plasmid by electroporation(Amaxa, Gaithersburg, MD). COS-7 cells were transiently trans-fected with the NY-ESO-1 encoding plasmid and/or the appropri-ate HLA class I expressing plasmids by using FuGENE (RocheDiagnostics, Basel, Switzerland). NA8-MEL and HT1080 tumorcell lines were transfected with an HLA-B35 expressing plasmid byelectroporation (Amaxa). mDCs, COS-7 cells, or tumor cell lines(105 per well) were incubated with the enriched T cell populations(2 � 104 per well) for 24 h, and IFN-� concentration was measuredin the culture supernatant by using ELISA as described in ref. 14.

We thank Rose-Marie Holman and Sean Lemoine for regulatory and datamanagement support. We are grateful to Coley Pharmaceutical Group(Wellesley, MA) for providing the CpG 7909. This study was supported bythe Cancer Vaccine Collaborative program of the Ludwig Institute forCancer Research and the Cancer Research Institute (New York, NY).

1. Scanlan MJ, Gure AO, Jungbluth AA, Old LJ, Chen YT (2002) Immunol Rev188:22–32.

2. Simpson AJ, Caballero OL, Jungbluth A, Chen YT, Old LJ (2005) Nat RevCancer 5:615–625.

3. Chen YT, Scanlan MJ, Sahin U, Tureci O, Gure AO, Tsang SL, Williamson B,Stockert E, Pfreundschuh M, Old LJ (1997) Proc Natl Acad Sci USA 94:1914–1918.

4. Carbone FR, Kurts C, Bennett SR, Miller JF, Heath WR (1998) ImmunolToday 19:368–373.

5. Ochsenbein AF, Sierro S, Odermatt B, Pericin M, Karrer U, Hermans J,Hemmi S, Hengartner H, Zinkernagel RM (2001) Nature 411:1058–1064.

6. Krieg AM (2003) Nat Med 9:831–835.7. Chu RS, Targoni OS, Krieg AM, Lehmann PV, Harding CV (1997) J Exp Med

186:1623–1631.8. Maurer T, Heit A, Hochrein H, Ampenberger F, O’Keeffe M, Bauer S, Lipford

GB, Vabulas RM, Wagner H (2002) Eur J Immunol 32:2356–2364.9. Heit A, Huster KM, Schmitz F, Schiemann M, Busch DH, Wagner H (2004)

J Immunol 172:1501–1507.10. Shinohara ML, Lu L, Bu J, Werneck MB, Kobayashi KS, Glimcher LH, Cantor

H (2006) Nat Immunol 7:498–506.11. Krieg AM (2002) Annu Rev Immunol 20:709–760.12. Valmori D, Souleimanian NE, Hesdorffer CS, Ritter G, Old LJ, Ayyoub M

(2005) Clin Immunol 117:24–30.13. Zeng G, Aldridge ME, Wang Y, Pantuck AJ, Wang AY, Liu YX, Han Y, Yuan

YH, Robbins PF, Dubinett SM, et al. (2005) Int J Cancer 114:268–273.14. Ayyoub M, Souleimanian NE, Godefroy E, Scotto L, Hesdorffer CS, Old LJ,

Valmori D (2006) Clin Immunol 118:188–194.15. Valmori D, Levy F, Godefroy E, Scotto L, Souleimanian NE, Karbach J, Tosello

V, Hesdorffer CS, Old LJ, Jager E, et al. (2007) Clin Immunol 122:163–172.16. Zarour HM, Maillere B, Brusic V, Coval K, Williams E, Pouvelle-Moratille S,

Castelli F, Land S, Bennouna J, Logan T, et al. (2002) Cancer Res 62:213–218.17. Davis ID, Chen W, Jackson H, Parente P, Shackleton M, Hopkins W, Chen Q,

Dimopoulos N, Luke T, Murphy R, et al. (2004) Proc Natl Acad Sci USA101:10697–10702.

18. Jager E, Karbach J, Gnjatic S, Neumann A, Bender A, Valmori D, Ayyoub M,Ritter E, Ritter G, Jager D, et al. (2006) Proc Natl Acad Sci USA 103:14453–14458.

19. Sallusto F, Lenig D, Forster R, Lipp M, Lanzavecchia A (1999) Nature401:708–712.

20. Ruprecht CR, Lanzavecchia A (2006) Eur J Immunol 36:810–816.21. Kumar S, Jones TR, Oakley MS, Zheng H, Kuppusamy SP, Taye A, Krieg AM,

Stowers AW, Kaslow DC, Hoffman SL (2004) Infect Immun 72:949–957.22. Papadea C, Check IJ (1989) Crit Rev Clin Lab Sci 27:27–58.23. Wille-Reece U, Flynn BJ, Lore K, Koup RA, Miles AP, Saul A, Kedl RM,

Mattapallil JJ, Weiss WR, Roederer M, et al. (2006) J Exp Med 203:1249–1258.24. Rothenfusser S, Hornung V, Ayyoub M, Britsch S, Towarowski A, Krug A,

Sarris A, Lubenow N, Speiser D, Endres S, et al. (2004) Blood 103:2162–2169.25. Speiser DE, Lienard D, Rufer N, Rubio-Godoy V, Rimoldi D, Lejeune F, Krieg

AM, Cerottini JC, Romero P (2005) J Clin Invest 115:739–746.26. Regnault A, Lankar D, Lacabanne V, Rodriguez A, Thery C, Rescigno M,

Saito T, Verbeek S, Bonnerot C, Ricciardi-Castagnoli P, et al. (1999) J Exp Med189:371–380.

27. Nagata Y, Ono S, Matsuo M, Gnjatic S, Valmori D, Ritter G, Garrett W, OldLJ, Mellman I, You Z, et al. (2002) Proc Natl Acad Sci USA 99:10629–10634.

28. Schnurr M, Chen Q, Shin A, Chen W, Toy T, Jenderek C, Green S, MiloradovicL, Drane D, Davis ID, et al. (2005) Blood 105:2465–2472.

29. Matsuo M, Nagata Y, Sato E, Atanackovic D, Valmori D, Chen YT, Ritter G,Mellman I, Old LJ, Gnjatic S (2004) Proc Natl Acad Sci USA 101:14467–14472.

30. Hulett MD, Hogarth PM (1994) Adv Immunol 57:1–127.31. van de Winkel JG, Capel PJ (1993) Immunol Today 14:215–221.32. Houghton AN, Scheinberg DA (2000) Nat Med 6:373–374.33. Hara I, Takechi Y, Houghton AN (1995) J Exp Med 182:1609–1614.34. Clynes R, Takechi Y, Moroi Y, Houghton A, Ravetch JV (1998) Proc Natl Acad

Sci USA 95:652–656.35. Dyall R, Vasovic LV, Clynes RA, Nikolic-Zugic J (1999) Eur J Immunol

29:30–37.36. Dhodapkar KM, Krasovsky J, Williamson B, Dhodapkar MV (2002) J Exp Med

195:125–133.37. Ittah M, Miceli-Richard C, Gottenberg JE, Lavie F, Lazure T, Ba N, Sellam

J, Lepajolec C, Mariette X (2006) Arthritis Res Ther 8:R51.38. Hartmann E, Wollenberg B, Rothenfusser S, Wagner M, Wellisch D, Mack B,

Giese T, Gires O, Endres S, Hartmann G (2003) Cancer Res 63:6478–6487.39. Stockert E, Jager E, Chen YT, Scanlan MJ, Gout I, Karbach J, Arand M, Knuth

A, Old LJ (1998) J Exp Med 187:1349–1354.

8952 � www.pnas.org�cgi�doi�10.1073�pnas.0703395104 Valmori et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

9, 2

020