A Gene Encoding an Antigen Recognized by
`Cytolytic T Lymphocytes on a Human Melanoma
`P. VAN DER BRUGGEN, C. TRAVERSARI,* P. CHOMEZ, C. LURQUIN,
`E. DE PLAEN, B. VAN DEN EYNDE, A. KNUTH, T. BOONt
`
`Many human melanoma tumors express antigens that are recognized in vitro by
`cytolytic T lymphocytes (CTLs) derived from the tumor-bearing patient. A gene was
`identified that directed the expression of antigen MZ2-E on a human melanoma cell
`line. This gene shows no similarity to known sequences and belongs to a family of at
`least three genes. It is expressed by the original melanoma cells, other melanoma cell
`lines, and by some tumor cells of other histological types. No expression was observed
`in a panel of normal tissues. Antigen MZ2-E appears to be presented by HLA-A1;
`anti-MZ2-E CTLs of the original patient recognized two melanoma cell lines of other
`HLA-Al patients that expressed the gene. Thus, precisely targeted immunotherapy
`directed against antigen MZ2-E could be provided to individuals identified by HLA
`typing and analysis of the RNA of a small tumor sample.
`
`resistant transfectants were obtained for
`each group and they were divided into mi-
`crocultures of 30 independent transfectants.
`These microcultures were allowed to ex-
`pand, duplicated, and tested for their ability
`to stimulate tumor necrosis factor (TNF)
`release by anti-E CITL clone 82/30 (11).
`This enables identification of those pools
`that contain one transfectant that expresses
`antigen E (8). Two of the 14 groups of
`cosmids produced positive microcultures,
`and the E' transfectants were recovered
`from the duplicates. After additional trans-
`fection experiments, a total of five E' trans-
`fectants out of 29,000 independent geneti-
`transfectants were obtained
`cin-resistant
`with the first positive group of cosmids and
`two out of 13,000 with the second group.
`By packaging the DNA of cosmid trans-
`fectants directly into lambda phage compo-
`nents, it is sometimes possible to retrieve
`cosmids that contain the sequences of inter-
`est (7, 12). This procedure was unsuccessful
`here, so we rescued the transfected sequence
`by ligating DNA of the transfectant to ap-
`propriate restriction fragments of cosmid
`vector pTL6 (13). This was tried with two
`transfectants and was successful with trans-
`fectant 7.4: cosmid B3 was obtained, which
`transferred the expression of antigen E at
`high efficiency. Transfectants obtained with
`this cosmid were as sensitive to lysis by the
`anti-E CTLs as the original melanoma cell
`line (Fig. 1).
`Cosmid B3 was digested with Sma I and a
`12-kb fragment could transfer the expres-
`sion of the antigen (Fig. 1). After the frag-
`ment was digested with Bam HI, a 2.4-kb
`fragment was found to transfer the expres-
`sion of antigen E at high efficiency (Fig. 1).
`This small fragment was used as a hybrid-
`ization probe on a Southern (DNA) blot
`prepared with Bamr HI-digested DNA of an
`
`enabled the isolation of genes for antigens
`recognized by CTLs on mouse tumors (7).
`This approach is based on the transfection of
`cosmid libraries prepared with DNA of cells
`that express the relevant antigen. Transfec-
`tants expressing the antigen are identified by
`their ability to stimulate the appropriate
`CTLs. As a first step, we tried to obtain
`transfectants expressing antigen MZ2-E (E)
`with genomic DNA of an E' MZ2-MEL
`subline. The DNA was transfected into E-
`antigen-loss variant MZ2-MEL.2.2, which
`had been obtained by selection with an
`anti-E CTL clone. A transfectant was ob-
`tained that was lysed by autologous anti-E
`CTL clone 82/30 (8).
`of 700,000
`A library
`independent
`cosmids was prepared with DNA of an E'
`MZ2-MEL subline and groups of 50,000
`cosmids were amplified (9). DNA from each
`group of cosmids was cotransfected into E-
`line MZ2-MEL.2.2 together with plasmid
`pSVtkneo3, that confers resistance to ge-
`neticin (10). Approximately 5000 geneticin-
`
`M OST MOUSE TUMORS EXPRESS AN-
`tigens that constitute potential
`targets for rejection responses in
`syngeneic hosts (1). Against some of these
`tumors, highly active and specific CGLs can
`be derived from immunized animals by re-
`stimulation in vitro with tumor cells (2).
`That the antigens recognized by these CTLs
`in vitro can be effective tumor-rejection an-
`tigens is indicated by the finding that tumor
`cells that had escaped immune rejection in
`vivo were found to be resistant to the tu-
`mor-specific CTLs (3).
`For human tumors, autologous mixed
`cultures of tumor cells and lymphocytes can
`generate CTLs that lyse the tumor cells (4).
`These anti-tumor CITLs do not lyse targets
`of natural killer cells and autologous control
`cells such as fibroblasts or EBV-transformed
`B lymphocytes. However, it is difficult to
`evaluate to what extent the antigens recog-
`by
`autolo-
`nized on human tumors
`gous CILs are relevant for tumor rejection.
`We have obtained a panel of autologous
`CTL clones (5) that lyse melanoma cell line
`MZ2-MEL, which was derived from patient
`MZ2. By selecting clonal sublines of MZ2-
`MEL that are not killed by these CTL
`clones, we obtained antigen-loss variants
`that were resistant to subsets of the CTL
`autologous
`demonstrating
`that
`clones,
`CITLs recognize a total of six independent
`antigens on MZ2-MEL (6). We then at-
`tempted to identify the gene for one ofthese
`antigens, MZ2-E, by an approach that had
`
`P. van der Bruggen, C. Traversari, P. Chomez, C.
`Lurquin, E. De Plaen, B. Van den Eynde, and T. Boon,
`Ludwig Institute for Cancer Research, B-1200 Brussels,
`Universitd
`Belgium
`Unit,
`Cellular
`Genetics
`and
`Carholique de Louvain, Brussels, Belgium.
`A. Knuth, I. Medizinische Klinik und Polilnik Jo-
`hannes Gutenberg-Universitat Mainz, D-6500 Mainz,
`Germany.
`
`*Present address: Ospedale S. Raffaele (Ematologia)
`20132 Milano, Italia.
`tTo whom correspondence should be addressed.
`
`13 DECEMBER 1991
`
`Targets
`E* tranafecbant obtaine
`wIth
`2 M11443-MEL
`Frament
`Fragmen
`3
`1S12kb
`DamM 2.4 kb
`
`_
`
`_
`
`MZ2-MEL.3.0
`E'
`
`M2M2EL
`
`E-
`
`~CoWmid Of
`group?7
`
`a0.
`eo0
`40.
`
`20
`
`10
`0
`
`in
`
`0-
`
`.1
`
`1
`
`.3
`
`3
`
`.11
`'
`'1 '1
`.3
`3
`
`.1 1
`1
`.1
`.3
`
`1
`
`.1
`
`1
`
`.3
`
`3
`
`.1
`
`1
`
`.3
`
`3
`
`.1
`
`1
`
`.3
`
`3
`
`.1
`'1'
`'
`.3
`3
`3
`Efferlarget ratio
`Fig. 1. Sensitivity of `5Cr-labeled target cells to lysis by anti-E CTL clone MZ2-CTL 82/30. Chromium
`release was measured after 4 hours (5). MZ2-MEL.3.0 is a clonal subline of the melanoma cell line of
`patient MZ2, and MZ2-MEL.2.2 is an antigen-loss variant selected with an anti-E CTL clone (6). The
`E' transfectant clones were isolated from MEL.2.2 cell populations transfected with cosmid group 7
`(transfectant 7.4); with cosmid B3; with a cloned 12-kb Sma I fragment of B3 (20); with a doned
`2.4-kb Bam HI fragment of the 12-kb fragment (20). Melanoma cell line MI13443-MEL was derived
`from another HLA-A1 patient.
`
`REPORTS
`
`1643
`
`Downloaded from https://www.science.org at REPRINTS DESK on October 03, 2024
`
`JHU 2081
`Merck Sharp v. Johns Hopkins
`IPR2024-00647
`
`1
`
`

`

`E' subline of MZ2-MEL and of the E-
`antigen-loss variant. The expected 2.4-kb
`band was observed only with the DNA of
`the E' melanoma cell (Fig. 2), indicating
`that the E- variant had lost the expression of
`the antigen as a result of a deletion affecting
`the relevant gene. Partial or complete dele-
`tions of tumor-antigen genes also occur in
`antigen-loss variants of mouse tumor cells
`(14). In addition to the 2.4-kb band, the
`probe hybridized to several additional bands
`of different intensities, suggesting that the
`gene responsible for the production of anti-
`gen E has sequence similarities to several
`other genes.
`The sequence of the transfecting 2.4-kb
`genomic segment showed no significant
`similarity to any sequence presently record-
`ed in data banks (15). Northern (RNA)
`blots and a cDNA library were prepared
`with RNA of E' subline MZ2-MEL.3.0
`(16). The 2.4-kb segment hybridized to an
`mRNA ofapproximately 1.8 kb on a North-
`ern blot. cDNA clones were obtained whose
`sequences were identical to parts of the
`2.4-kb genomic fragment, thereby identify-
`ing two exons in this fragment. The position
`of one additional exon located upstream was
`obtained by sequencing segments of cosmid
`B3 that were located in front of the 2.4-kb
`Bam HI fragment. The gene extends over
`approximately 4.5 kb (Fig. 3). The starting
`point of the transcribed region was con-
`firmed by polymerase chain reaction (PCR)
`amplification of the 5' end of the cDNA
`(17). The three exons are 65, 73, and 1551
`bp, respectively (Fig. 3). An ATG located in
`position 66 of exon 3 is followed by an open
`reading frame of 828 bp.
`The ability ofthe 2.4-kb gene fragment to
`transfer the expression ofantigen E confirms
`
`00
`
`Ul
`C'2
`
`-W
`
`N
`
`N
`
`Ula
`
`W
`
`N
`
`kb
`-12
`
`-8
`
`- 6
`
`-4
`
`3
`
`-2
`
`.1
`
`Fig. 2. Identification ofa
`genomic deletion in the
`E- antigen-loss variant.
`The 2.4-kb Bam HI
`fragment, which trans-
`ferred the expression of
`antigen MZ2-E, was la-
`beled with 32P and used
`as a probe on a Southern
`blot of Bam HI-digest-
`ed DNA of E' clonal
`subline MZ2-MEL3.0
`and of E- variant MZ2-
`MEL.2.2. The 2.4-kb
`band is absent in the lane
`of the E- variant. The
`DNA of CTL clone 82/
`30 of patient MZ2 dis-
`played the same bands as
`the E' melanoma cells.
`
`Sma I rf
`
`Bam HI
`
`1
`
`0
`
`1
`
`2
`
`3
`
`4
`
`5kb
`
`Fragments transferring the
`expression of antigen E
`
`2,Akb
`
`912
`
`215
`255
`Fig. 3. Structure of the gene of antigen MZ2-E
`and restriction sites. Exons are indicated as black
`boxes and the open reading frame in exon 3 is
`marked in white. Boundaries of transfecting frag-
`ments are indicated relative to the first nucleotide
`of exon 3.
`
`previous observations that truncated genes
`lacking the promoter can efficiently transfer
`CTL epitopes (7). Smaller regions of the
`gene corresponding to parts of the 2.4-kb
`fragment were cloned and tested by trans-
`fection into E- cells. Three segments trans-
`ferred the expression of the antigen (Fig. 3).
`Thus, the gene probably encodes the anti-
`genic peptide recognized by the anti-E
`CTLs, as opposed to producing a protein
`that activates the encoding gene. The encod-
`ing sequence would be where all transfecting
`fragments overlap, as is the case for the four
`mouse tumor antigens that we have studied.
`Two nonidentical cDNA species were
`also found when the cDNA library was
`probed with the 2.4-kb fragment. This con-
`firmed the existence of a gene family sug-
`gested by the pattern observed on the
`Southern blots (Fig. 2). In contrast with the
`first cDNA, the second and the third types
`ofcDNA were unable to transfer the expres-
`sion of antigen E in transfection experi-
`ments. No significant homology was found
`by comparing the sequences of the two
`additional cDNAs to those presently record-
`ed in gene banks (15). We propose the name
`"MAGE" (melanoma antigen) for this new
`gene family, with MAGE-1 referring to the
`
`gene that directs the expression of antigen
`MZ2-E and MAGE-2 and -3 for the two
`other genes (Fig. 4). Analysis of the third
`exon showed that the two additional cDNAs
`are more closely related to each other (12%
`differences) than to the first (18.1% and
`18.9% differences). Out of nine cDNA
`obtained with RNA of MZ2-
`clones
`MEL.3.0, three of each type were obtained,
`suggesting approximately equal expression
`of the three genes. It is possible that other
`closely related genes are expressed in lower
`amounts in these cells.
`Experiments with mouse tumors showed
`that new antigens recognized by T cells can
`result from point mutations in the coding
`region of active genes. New antigens can
`also arise from the activation of genes that
`are not expressed in most normal cells (7).
`To clarify this issue for antigen MZ2-E, we
`compared the MAGE-1 gene present in the
`melanoma cells to that present in normal
`cells of patient MZ2. We amplified by poly-
`merase chain reaction (PCR) the DNA of
`phytohemagglutinin-activated blood lym-
`phocytes with primers surrounding a 1300-
`bp stretch covering the first half of the
`2.4-kb fragment. A PCR product was ob-
`tained, whereas none was obtained with the
`DNA of the E- variant. The sequence of
`this PCR product was identical to the cor-
`responding sequence of the gene carried by
`the E' melanoma cells. Moreover, we found
`that antigen MZ2-E was expressed by cells
`transfected with the cloned PCR product.
`Thus, the activation of a normal gene is
`responsible for the appearance of antigen
`MZ2-E. One may wonder how E- antigen-
`loss variants could be obtained in these cir-
`cumstances, because both copies of the gene
`would have to be inactivated. The same phe-
`nomenon has been observed with homozy-
`gous mouse tumor P815 (3). One explana-
`tion would be the existence among the
`
`MAGE-
`
`MAGB-2
`
`MAGB-1
`
`xxx ccsvcccaos
`
`GTCCTCAGGG&GCCTCCAgCcTcCCACTACATgAACTaCcCTCtctg.G1jCtCCtaAGAGGacTCC.a CCaaaGA&GGGAGG
`C26-X
`GTCCTCAGGG&GCCTCCagCTTctCgACTACCATCAACTaCACTCttttgGA!CAtCACG
`TGAGGGcTCACACAGCCaaGA&GaGGAGG
`xx CCTCCCCp.cM
`xr
`
`225 cso-I
`DCTtccTgaCC-TGGAGTCCgaGTTCCAGCAcTCAgTAgGAGGTGCcGgTTGGTTc-----C-
`A CGG
`zzz GGMCA&CA
`tg~tTcccgaCCtTGGAGTCCGAGTTCCaAGCAGcAATCAgSTAgGAGaTGGtTGagTTGGTTca=rTCTAT G CA
`GGCCAAG&At
`SAGC&
`CTCTGTATcC-
`x GCCAAGA.
`
`325
`ZXt GGGGcgGs
`ZZ GGC'.GCC~
`Z GG3AG.CCAlG
`425
`
`GmGCTIGC
`
`rCACAA&GGC=_A&ATGCTG~gGAGTG~cgT~gAAATTggcAGtAtTtcT TTCCTGtGATCTTCaGCAA&GCtTCcagtnCTGCAGCT
`TCcTCAgAAATT9gCcAGgACTtcTTTCccGtGATcCTTCaGCAAAGCcAG
`rCA~aAGGC&GAAgATGCTGAG
`CTTGC&GCT
`5E0o-4
`ATcGAgcTGAtGGAAGtgGAcCCCAtCGGCCACTtgTAc&TCtTTGcC&CCTGCCTgGGcCCTCTCTAcG&GcTCGGGC&
`cCGcTg TcCTCG cCTGCTGCGACAAT
`XAcGAqGTGgtGGAAGtgGtcCCCAtCaGCCACTtgAca
`CAT$3AcGTAG5AAPAGtgcCcGGCC&CcC.AtaGCCTGC~tgScC _ _C_.
`
`XCCAAGCAGGCcTCCTGATAATcGTCCTGGcCATaATcGCAAgaGAGGGCtgTGCcG
`OCAAGACAQGC~cSCCGATAAT'IG~
`GAT-G~CA&TGGaG
`CATGCA
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`.
`
`525
`G CG
`CAGATCATGC
`CAGgcATGC
`GAtgTG~TCTGGGcTGAGTG
`CAGATCATGC
`. 0
`..
`625
`C2O-9
`Fig. 4. Part of exon 3 of gene MAGE-1, which directs the expression of antigen MZ2-E, and of related
`genes MAGE-2 and MAGE-3. Lowercase letters in the two latter sequences indicate differences relative
`to MAGE-1. Numbering is relative to the first nucleotide of exon 3. Oligonucleotides CHO-8 and
`CHO-9 were used to prime the reverse transcriptions and PCR amplification of Fig. 6. Oligonudeo-
`tides SEQ4, CHO-2, and CHO-3 were used as specific hybridization probes discriminating MAGE-1,
`-2, and -3. The codons of the open reading frame are indicated by points.
`
`1644
`
`SCIENCE, VOL. 254
`
`Downloaded from https://www.science.org at REPRINTS DESK on October 03, 2024
`
`2
`
`

`

`cultured cell populations of a subset of cells
`that are haploid for the relevant chromosome.
`To evaluate the expression ofMAGE-1 by
`various normal and tumor cells, we hybrid-
`ized Northern blots with a probe covering
`most of the third exon. In contrast with the
`result observed with tumor cell line MZ2-
`MEL.3.0, no band was observed with RNA
`isolated from a CTL clone of patient MZ2
`and phytohemagglutinin-activated
`blood
`lymphocytes of the same patient. Also neg-
`ative were several normal tissues of other
`individuals (Fig. 5 and Table 1). Ten of 14
`melanoma cell lines of other patients were
`positive to varying degrees. Two of four
`samples of melanoma tumor tissue, includ-
`
`CMNWl <o0
`
`° v
`
`I
`
`CW, N
`
`Other melanomas
`
`1 W
`
`t
`
`_~~~~~~~~~~~~m
`
`or tumors
`Othe
`
`.~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
`
`i,
`
`_,
`
`i,
`
`,
`
`W
`
`m
`
`_
`
`three MACE genes. The PCR products were
`;2, were pos-
`ing a metastasis of patient MZ
`then tested for their ability to hybridize to
`ssion of the
`itive, excluding that the expre
`three other oligonucleotides that showed
`ture artifact.
`gene represented a tissue cull
`complete specificity for one ofthe three genes
`logical types
`Some tumors of other histol
`(Fig. 4). Control experiments that were done
`id Table 1).
`were also positive (Fig. 5 an
`by diluting RNA of melanoma MZ2-
`is expressed
`Thus, the MAGE gene family
`by other tu-
`MEL.3.0 into RNA from negative cells indi-
`by many melanomas and also
`cated that under our conditions the intensity
`probe cross-
`mors. However, because the
`of the signal decreased proportionally to the
`indication as
`hybridizes, there was no clear
`dilution and that positive signals could still be
`to which of the three genes we
`are expressed
`c)rted to PCR detected at a dilution of 1/300. The normal
`by these cells. We therefore resc
`cells that were tested by PCR were confirmed
`with specific
`amplification and hybridization
`to be negative for the expression of the three
`oligonucleotide probes. cDNI
`ks were ob-
`vith oligonu- MAGE genes, suggesting a level of expres-
`tained and amplified by PCR v
`sion of <1/300 that of the MZ2 melanoma
`nded to se-
`cleotide primers that correspc
`:ntical for the
`cell line (Fig. 6). Some melanomas expressed
`quences of exon 3 that were ide
`MACE genes 1, 2, and 3 whereas others
`expressed only MAGE-2 and -3 (Fig. 6 and
`Table 1). Some of the other tumors also
`expressed all three genes, whereas others ex-
`pressed only MAGE-2 and -3 or only
`MAGE-3. The MAGE gene family, thus, is
`expressed by a large array of different tumors
`and not by most normal cells. The MACE
`genes may participate in tumor transforma-
`tion or in aspects of tumor progression such
`as the ability to metastasize. The observation
`that MACE-1 is expressed in several melano-
`mas shows that tumors of different patients
`can express the same tumor rejection antigen
`recognized by autologous CTLs.
`We also determined the associated major
`........... _;
`histocompatibility complex (MHC) class I
`MW -
`molecule. The class I specificities of patient
`Fig. 5. Northern blot analysis of the expression of MACE-1 in tumor cell lines, tumol
`a 1.3-kb DNA MZ2 are HLA-A1, A29, B37, B44, and C6.
`normal human tissues. All lanes contained 20 jig oftotal RNA and were hybridized with
`Four other melanomas of patients that had
`in probe (21).
`probe extending from positions 255 to 1544 of exon 3 of MAGE-1 and with a F-act
`Al in common with MZ2 were cotrans-
`Hybridization with both probes was performed successively on the same blot. Hyb
`oridization and
`washing conditions were the same for all the experiments.
`fected with the 2.4-kb fragment and pSVt-
`kneop. Three of them yielded neor transfec-
`tants that stimulated TNF release by anti-E
`CTL clone 82/30, which is CD8' (Table 1).
`No E' transfectant was obtained with four
`other melanomas, some of which shared
`A29, B44, or C6 with MZ2. Thus, the
`presenting molecule for antigen MZ2-E is
`probably HLA-Al. Eight melanomas of pa-
`tients with HLA haplotypes that did not
`include Al were examined for their sensitiv-
`ity to lysis and for their ability to stimulate
`TNF release by the CTLs. None were pos-
`itive even though three expressed MACE-1
`(Table 1). Out of six melanoma cell lines
`derived from tumors of HLA-Al patients,
`the two lines that expressed MAGE-1 also
`stimulated TNF release by anti-E CTL clone
`82/30 of patient MZ2. One of these tumor
`cell lines, M113443-MEL, also showed high
`sensitivity to lysis by these anti-E CTLs
`(Fig. 1). The ability of some human anti-
`tumor CTLs to lyse allogeneic tumors that
`share an appropriate HLA specificity with
`the original tumor has been reported (18).
`It is difficult to definitively establish that
`antigens recognized in vitro on human tu-
`
`Probes
`
`i
`
`W.
`
`W
`
`W
`
`mmm
`
`MAGE A
`~ ~
`normalhuman tissues.~~~
`Al ae otand2
`
`c A
`
`A
`
`ofttlKA n
`
`5m
`eehbrdzdwt
`
`"actinI:
`
`UJ
`
`CM
`
`NiON
`
`0 c
`
`'s
`
`UJ
`
`CM
`N
`2
`
`Probes
`
`MAGE-1
`
`MAGE-2
`
`MAGE-3
`
`Normal
`Other
`tissues Melanomas tumors
`
`(D~ ~
`-J _j W
`
`~
`
`-J
`
`~
`
`cc
`c
`
`Normal
`tissues Melanomas Other tumors
`~
`C')
`W
`2
`.,
`W W -i _W
`
`W
`qC' 1WU )_
`w
`0 0
`jcJ 00
`mN m
`2 2 m m X5 >3
`
`,-X x)
`
`Dm mm - m
`-i -in
`
`'>= %.
`N
`nX> 2 2 > -i m >1
`N
`
`C) a a
`Q
`cj
`
`Fig. 6. Detection of the expression of MAGE-1 and of related genes MAGE-2 and -3 by reverse
`transcription and PCR amplification. Total RNA was extracted from the tumor cell lines, tumor
`samples, and normal human tissues. Oligonucleotides CHO-8 and CHO-9 (Fig. 4), which correspond
`to identical regions in MAGE- 1, -2, and -3 were used to prime cDNA synthesis and PCR amplification
`(22). The PCR products were fractionated by size in agarose gels and blotted on nitrocellulose. The
`blots were hybridized with oligonucleotides probes SEQ-4, CHO-2, and CHO-3, which are highly
`specific for sequences ofMAGE-1, -2, and -3, respectively (Fig. 4). This specificity is demonstrated in
`the right part of the figure showing hybridization with these three probes of PCR-amplified MAGE- 1,
`-2, and -3 cDNA clones. The nitrocellulose filters corresponding to both panels were hybridized,
`washed, and autoradiographed together.
`
`13 DECEMBER 1991
`
`REPORTS
`
`1645
`
`Downloaded from https://www.science.org at REPRINTS DESK on October 03, 2024
`
`3
`
`

`

`mor cells by autologous CILs constitute po-
`tential targets for immune rejection in vivo.
`This relevance has been demonstrated for
`mouse tumor antigen P81SA (3), and we find
`a striking similarity between this mouse tu-
`mor rejection antigen and antigen E of hu-
`man melanoma MZ2-MEL. Both antigens
`are on tumors that express four to six antigens
`recognized by autologous CTLs (3, 6). The
`genes coding for both antigens appear to be
`silent or quasi-silent on most normal tissues
`and are activated in the tumors. The se-
`quences of both genes appear to be identical
`in normal tissues and in the tumors. Finally,
`both genes are expressed in several indepen-
`dent tumors, resulting in the expression ofthe
`antigen in those tumors that carry the appro-
`priate class I MHC molecule.
`
`The finding that a potential tumor rejec-
`tion antigen is shared by a significant pro-
`portion of human tumors and the ability to
`identify these tumors readily on the basis of
`their expression of the relevant gene may
`have important implications for cancer im-
`munotherapy.
`Small tumor samples of
`HLA-A1 individuals [26% of total in Cau-
`casian populations (19)] could be frozen and
`the RNA tested by reverse transcription and
`PCR amplification to identify the tumors
`that express MAGE-1. These tumors should
`express antigen E and may therefore be
`sensitive to an anti-E T lymphocyte re-
`sponse. Additional study will be required to
`assess how much MAGE-1 expression is
`needed for effective production of antigenic
`peptides as well as the critical density ofclass
`
`I molecules and adhesion molecules re-
`quired at the cell surface. But eventually this
`should lead to precisely targeted strategies of
`active or passive immunization. A promising
`possibility for active immunization involves
`the use of cells engineered for high expres-
`sion of MAGE-1, HIA-Al, and possibly
`genes coding for some adhesion molecules
`and interleukins.
`
`REFERENCES AND NOTES
`1. R. T. Prehn and J. M. Main, J. Nad. Cancer Inst.
`18, 769 (1957); G. Klein, H. Sjogren, K. E.
`Hellstrom, Cancer Res. 20, 1561 (1960); A. Van
`Pel, F. Vessiere, T. Boon, J. Exp. Med. 157, 1992
`(1983); E. Fearon, T. Itaya, B. Hunt, B. Vogelstein,
`P. Frost, Cancer Res. 48, 2975 (1988).
`2. K. Brunner, R. MacDonald, J. C. Cerottini, J. Exp.
`Med. 154, 362 (1981); T. Boon, J. Van Snick, A.
`
`Table 1. Expression of genes MAGE-1, -2, and -3 and of antigen MZ2-E by tumors and normal tissues.
`
`Sample
`
`Melanoma cell line MZ2-MEL.3.0
`Tumor sample MZ2 (1982)
`Antigen-loss variant MZ2-MEL.2.2
`CTL clone MZ2-CTL.82/30
`PHA-activated blood lymphocytes
`
`Liver
`Muscle
`Skin
`Lung
`Brain
`Kidney
`
`LB34-MEL
`MI665/2-MEL
`MI10221-MEL
`M113443-MEL
`SK33-MEL
`SK23-MEL
`
`LB17-MEL
`LB33-MEL
`LB4-MEL
`LB41-MEL
`MI4024-MEL
`SK29-MEL
`MZ3-MEL
`MZ5-MEL
`
`BB5-MEL
`
`Small cell lung cancer H209
`Small cell lung cancer H345
`Small cell lung cancer H510
`Small cell lung cancer LB11
`Bronchial squamous cell carcinoma LB37
`Thyroid medullary carcinoma IT
`Colon carcinoma LB31
`Colon carcinoma LS411
`
`Expression ofMAGE gene family
`cDNA-PCR't
`MAGE-2
`
`MAGE-1
`
`Norer
`Northe_
`MAGE-1 probe*
`
`Recognition by
`E-specific CTL
`_transfection¶
`MAGE-3
`TNFt
`Lysis§
`
`E + after
`
`Cells ofpatient MZ2
`++++
`+++
`-
`
`++++
`+++
`+ + +
`
`++++
`+++
`+ + +
`
`+
`
`+
`
`Normal tissues
`
`+
`+
`
`_
`+
`
`_
`_
`
`+
`_
`-
`+
`-
`-
`
`-
`-
`
`-
`
`_
`_
`
`_
`-
`+
`-
`-
`
`-
`-
`
`-
`
`_
`_
`
`Melanoma cell lines ofHLA-Al patients
`++
`++++
`-
`-
`-
`+ +
`++++
`+++
`-
`++++
`-
`++++
`Melanoma cell lines ofother patients
`+
`++++
`-
`+++
`
`+++
`
`++++
`
`+
`++++
`-
`++++
`Melanoma tumor sample
`+++
`++
`Other tumor cell lines
`-
`-
`-
`+
`-
`++++
`-
`
`++++
`++++
`++++
`++++
`-
`+++
`+++
`
`++++
`_
`
`+
`++++
`++++
`++++
`
`+++
`+++
`
`++++
`
`++++
`++++
`
`+++
`
`++++
`++++
`++++
`++++
`+++
`++++
`++++
`
`+
`+
`+
`
`-
`
`+
`
`+
`+
`+
`+
`
`+
`+
`
`+
`
`+
`+
`
`+
`
`+
`+
`+
`+
`+
`+
`+
`
`Chronic myeloid leukemia LLC5
`Acute myeloid leukemia TA
`tTNF release by CTL 82/30 after stimulation with
`*Data obtained in the conditions of Fig. 5 with a cross-reactive MAGE-1 probe.
`tData obtained as described in Fig. 6.
`1Ceils transfected with the 2.4-kb fragment of gene MAGE-I
`SLysis of5`Cr-labeled target by CIL 82/30 in the conditions of Fig. 1.
`the tumor cells as described in (11).
`were tested for their ability to stimulate TNF release by CIL 82/30.
`
`-
`
`Other tumor samples
`-
`
`-
`
`-
`
`1646
`
`SCIENCE, VOL. 254
`
`Downloaded from https://www.science.org at REPRINTS DESK on October 03, 2024
`
`4
`
`

`

`Van Pel, C. Uyttenhove, M. Marchand, ibid. 152,
`1184 (1980); C. J. M. Melief, Adv. Cancer Res. 58,
`143 (1991).
`3. C. Uyttenhove, J. Maryanski, T. Boon, J. Exp.
`Med. 157, 1040 (1983).
`4. A. Anichini, G. Fossati, G. Parmiani, Immunol.
`Today 8, 385 (1987).
`5. M. Herin et al., Int.J. Cancer 39, 390 (1987).
`6. B. Van den Eynde et al., ibid. 44, 634 (1989).
`7. T. Wolfel et al., Immunogenetics 26, 178 (1987); E.
`De Plaen et al., Proc. Natl. Acad. Sci. U.SA. 85,
`2274 (1988); C. Lurquin et al., Cell 58, 293
`(1989); J. P. Szikoraetal., EMBOJ. 9, 104 (1990);
`C. Sibille et al., J. Exp. Med. 172, 35 (1990); B.
`Van den Eynde, B. Lethe, A. Van Pel, E. De Plaen,
`T. Boon, ibid. 173, 1373 (1991).
`8. C. Traversari et al., Immunogenetics, in press.
`9. The genomic library was constructed with DNA
`from MZ2-MEL.43 as described in E. De Plaen
`et al., Proc. Nati. Acad. Sci. U.S.A. 85, 2274
`(1988).
`10. We used the transfection procedure described in (8)
`with the following modification: 4.5 x 106 cells
`attached to 600 cm2 tissue culture flasks (Singletray
`Unit, Nunc) containing 180 ml of medium were
`treated with a 20-mi calcium phosphate-DNA pre-
`cipitate of 240 jig of cosmid DNA and 24 pLg of
`pSVtkneop.
`11. As described in (8). Briefly, 1,500 CTI
`specific for E
`were added to 4 x 104 transfected cells in a micro-
`well. After 24 hours, 50 gIl of the supernatant were
`added to 3 x 104 cells of cell line WEHI 164 clone
`13. This TNF-sensitive cell line was developed by T.
`Espevik and J. Nissen [J. Immunol. Methods 95, 99
`(1986)]. The mortality ofWEHI cells was estimated
`24 hours later by a colorimetric assay described by
`M. B. Hansen, S. E. Nielsen, and K. Berg [ibid. 119,
`203 (1989)].
`12. Y. F. Lau and Y. W. Kan, Proc. Natd. Acad. Sci.
`U.SA. 80, 5225 (1983).
`13. T. Lund, F. G. Grosveld, R. A. Flavell, ibid. 79, 520
`(1982); E. De Plaen et al., ibid. 85, 2274 (1988).
`14. C. Lurquin et al., Cell 58, 293 (1989).
`15. The computer research for sequence homology was
`done with GenBank release 68 and the FASTA
`program, described by W. R. Pearson and D. J.
`Lipman [Proc. Natt. Acad. Sci. U.SA. 85, 2444
`(1988)]. The accession number of MAGE-1 in
`GenBank is M77481.
`16. cDNA libraries in bacteriophage XgtlO were pre-
`pared with the Amersham cDNA synthesis and
`cloning kits.
`17. Amplification of the 5' end of the cDNA by PCR as
`described in M. Frohman, M. Dush, G. Martin,
`Proc. Nat. Acad. Sci. U.S.A. 85, 8998 (1988). The
`primer for the synthesis of the cDNA was 5'-
`TTGCCGAAGATCTCAGGAA-3'. For the ampli-
`fication, we used as 3' primer the oligonucleotide
`5'-CTTGCCTCCTCACAGAG-3' and the 5' prim-
`ers described by Frohman et al.
`18. T. L. Darrow, C. L. Slingluff, Jr., H. F. Seigler, J.
`Immunol. 142, 3329 (1989); S. F. Slovin, R. D.
`Lackman, S. Ferrone, P. E. Kiely, M. J. Mastran-
`gelo, ibid. 137, 3042 (1986); N. J. Crowley, C. L.
`Slingluff, Jr., T. L. Darrow, H. F. Seigler, Cancer
`Res. 50, 492 (1990).
`19. J. L. Tiwari and P. I. Terasaki, in HLA and Disease
`Associations, T. L. Tiwari and P. I. Terasaki, Eds.
`(Springer-Verlag, New York, 1985), pp. 4-17.
`20. pTZ18R and pTZL9R were used as cloning vectors
`in order to produce double-stranded DNA for trans-
`fection and single-stranded DNA for sequencing.
`21. RNA isolation was performed as in L. G. Davis, M.
`D. Dibner, J. F. Battey, Basic Methods in Molecular
`Biology (Elsevier, New York, 1986), pp. 130-135.
`Northern blot analysis was performed as by B. Van
`den Eynde et al. in (7).
`22. For cDNA synthesis, total RNA (1 jig) was diluted to
`a total volume of 20 I. with 2 p1 of 10x buffer from
`the GENEAmp kit (Perkin Elmer-Cetus), 2 g1 each of
`10 mM dNTP, 1.2 gI of 25 mM MgCI2, 1
`i.1 of a
`80-pM solution of oligonudeotide primer CHO-9,
`20 units of RNAsin (Promega), and 200 units of
`MoMLV reverse transcriptase (BRL). This mixture
`was incubated at 42'C for 40 min. For PCR amplifi-
`cation, 8 p.1 of 10x buffer, 4.8 p.1 of 25 mM MgCl2,
`1-ptl of a 80-pLM solution ofprimer CHO-8, 2.5 units
`
`13 DECEMBER 1991
`
`of Taq polymerase, and water were added to a total
`volume of 100 pl.. Amplification was performed for
`30 cycles (1 min at 94TC, 2 min at 52C, and 3 min at
`72TC). Each reaction (10 pl) was size-fractionated in
`agarose gels, blotted on nitrocellulose, and hybridized
`with 52P-labeled oligonucleotides. Hybridization and
`washing conditions were as in (14).
`23. We hank P. Coulie and F. Lehmann fbr contributing
`to CIL lysis assays; P. Hainaut, P. Weynants, F.
`Brasseur, G. Parniani, L. Old, E. Stockert, and L.
`
`Suardet for providing tumor cell lines; F. Lejeune for
`melanoma tumor samples; A. M. Lebacq, J. J. Cassi-
`man, P. Marijnen, and E. Bellefroid for RNAofnormal
`and tumor cells; A. Authom for technical assistance;
`and S. Khaoulali for secretarial assistance. C.T. is sup-
`ported by a postgraduate fellowship for Cancer Re-
`search Training of the Commission of the European
`Communities.
`
`16 September 1991; accepted 31 October 1991
`
`Cell-Free, De Novo Synthesis of Poliovirus
`AKHTERuzzAMAN MOLLA,* ANIKO V. PAUL, EcKARD WIMMER
`
`Cell-free translation ofpoliovirus RNA in an extract ofuninfected human (HeLa) cells
`yielded viral proteins through proteolysis of the polyprotein. In the extract, newly
`synthesized proteins catalysed poliovirus-specific RNA synthesis, and formed infec-
`tious poliovirus de novo. Newly formed virions were neutralized by type-specific
`antiserum, and infection of human cells with them was prevented by poliovirus
`receptor-specific antibodies. Poliovirus synthesis was increased nearly 70-fold when
`nucleoside triphosphates were added, but it was abolished in the presence of inhibitors
`of translation or viral genome replication. The ability to conduct cell-free synthesis of
`poliovirus will aid in the study of picornavirus proliferation and in the search for the
`control of picornaviral disease.
`
`V IRUSES ARE COMPLEX AGGREGATES
`of organic macromolecules that as-
`semble from multiples of either a
`few or many distinct building blocks (1).
`Cell-free morphogenesis of infectious virus
`particles, in contrast to de novo synthesis,
`has been reported with some plant viruses
`and bacteriophages (2). In these cases, the
`viral structural components were isolated
`from virions or from infected cells. All pre-
`viral attempts to assemble animal viruses,
`including members of the Picornaviridae
`family, have been unsuccessful.
`The prototype picornavirus, poliovirus, is
`a non-enveloped, icosahedral particle con-
`sisting of a single-stranded RNA genome
`that is surrounded by 60 copies each of
`capsid polypeptides VP1 and VP3, 58 to 59
`copies each of VP2 and VP4, and one to
`two copies of VP0, the precursor to VP2
`and VP4 (3-5). Poliovirion formation in-
`volves the proteolysis of the capsid polypep-
`tide precursor P1 and the formation of
`"immature' protomers [(VPO,VP1, and
`VP3)5], procapsids (an empty shell consist-
`ing of 12 immature protomers) and, possi-
`bly, "provirions" (3). Provirions are nonin-
`fectious RNA-containing particles whose
`VPO polypeptides have not been cleaved to
`yield the mature virion (6). The formation
`of provirions in a cell-free extract of polio-
`virus-infected HeLa cells has been described
`but, for unknown reasons, these particles
`
`Department of Microbiology, School of Medicine, State
`University of New York at Stony Brook, Stony Brook,
`NY 11794.
`
`*To whom correspondence should be addressed.
`
`did not "mature" to virions (7). Palmenberg
`(8) and Grubman et al. (9) have succeeded in
`assembling capsid intermediate structures by
`translating, in rabbit reticulocyte lysates,
`RNAs of the picornaviruses encephalomyo-
`carditis virus and foot-and-mouth disease
`virus. Formation of infectious particles in
`these lysates was not reported, and it
`is
`conceivable that the rabbit reticulocyte ly-
`sate lacks components essential for morpho-
`genesis (10).
`Numerous problems
`concerning
`the
`structure and replication of picornaviruses
`remain unsolved. These include the function
`of virus-encoded polypeptides, the. mecha-
`nism of initiation of polyprotein synthesis,
`the mechanism of genome replica