A Nonimmunogenic Sarcoma Transduced with the
cDNA for Interferon 3/Elicits CD8 + T Cells
against the Wild-type Tumor: Correlation with
Antigen Presentation Capability
By Nicholas P. Restifo, Paul J. Spiess, Stephen E. Karp,
James J. Mul6, and Steven A. Kosenberg
From the SurgeryBranch, Divisionof Cancer Treatment,National CancerInstitute, National
Institutes of Health, Bethesda,Maryland20892
Summary
To be recognized by CD8 + T lymphocytes, target cellsmust process and present peptide antigens
in the context ofmajor histocompatibility complex(MHC) classI molecule~ The nonimmunogenic,
low classI-expressing, methylcholanthrene(MCA)-induced routine sarcoma cell line, MCA 101,
is a poor presenter of endogenously generated viral antigens to specific CD8 + T lymphocytes
and cannot be used to generate tumor infiltrating lymphocytes(TIL). Since interferon "y(IFN-3')
has been shown to upregulate three sets of molecules important for antigen processing and
presentation, we retrovirally transducedwild-type MCA 101 (101.WT) tumor with the mIFN-'y
eDNA to create the 101.NAT cell line. Unlike 101.WT, some clones of retrovirally transduced
101.NAT tumor expressed high levelsof class I, and could be used to generate CD8 + TIL. More
importantly, these TIL were therapeutic in vivo against established pulmonary metastases from
the wild-type tumor. Although not uniformly cytotoxic amongst severalseparate cultures, these
TIL did specificallyreleasecytokines (IFN-3, and tumor necrosis factor-c0 in response to 101.WT
targets. 101.WT's antigen'presentationddidt was also reversedby genemodificationwith mIFN-3,
eDNA. 101.NAT had a greatly improved capacity to present viral antigens to CD8 + cytotoxic
T lymphocytes. These findings show that a nonimmunogenic tumor, incapable of generating
a CD8 + T cell immune response, could be gene-modified to generate a therapeutically useful
immuneresponse against the wild-typetumor. This strategy maybe useful in developing treatments
for tumor histologies not thought to be susceptible to T cell-based immunotherapy.
Some tumor cells in mouse and humans present peptide/MHC class I complexes recognizable by T cells.
CD8 + CTL can be generated that specifically lyse human
and murine tumor cells in vitro and can eliminate established
tumor in vivo (1-12). Many tumors, however, are thought
to be nonimmunogenic (3, I3). Although there are many
possible explanations for nonimmunogenicity,it has recently
been demonstrated that a nonimmunogenic murine tumor
poorly presented endogenously generated viral antigens in
the context of MHC class I, despite the intracellular presence of these antigens in high quantities (14). Furthermore,
in a mutant lymphoma cell line with a known antigen presentation defect, tumor antigens were not presented, while
in its parent, nonmutated line they were (15). These findings
suggested that a tumor might evade recognition by CD8 +
T cells by failing to present tumor antigens that were not
absent, as Hewitt et al. hypothesized (13), but that were instead hidden intracelhlarly, and not presented on the cell
surface.
Our understandingof the molecular mechanisms that might
be involved in the evasion of recognition by CD8 + T cells
has grown significantly over the past several years. The
specificityof T cellrecognitionof a cognatepartner celloccurs
via the interaction of the TCR with a MHC molecule to
which a peptide is bound. Some peptide/MHC complexes
that are designated antigenic trigger a T cell response that
can consist ofproliferation, upregulation of surfacemolecules,
activation of lytic machinery, and/or secretion of cytokines.
The peptides presented by class I molecules are usually fragments of endogenous proteins that are cleaved, frequently
to an ultimate size of eight or nine amino acids (16-20). This
task is possibly achieved, in part, by a molecular complex
known as the proteasome (21, 22). Peptide fragments may
be transported across the membrane of the endoplasmicretic-1423 TheJournalof ExperimentalMedicine9Volume175 June1992 1423-1431
ulum (ER) t or some post-ER compartment by specialized
proteins, related to the products of the mnlti-drug resistance
genes, designated TAP 1 and 2 (23-26). Both the putative
peptide transporter proteins and the proteasome component
molecules appear to be very closely associated with the MHC
region on chromosome 17 in the mouse, or on chromosome
6 in the human (21-26). Thus, there may be an antigen presentation kit consisting of the three groups of molecules apparently required for endogenous antigen presentation: proteasome, peptide transporter, and class I H chain with
B2-microglobulin. It is significant that the known components of this equivalent of a eukaryotic operon for antigen
presentation are all upregnlatable by IFN-3r (27).
In the experiments reported here, we set out to study the
effects of insertion of the cDNA for mlFN-7 on antigen processing and on presentation in a tumor. The nonimmunogenic murine methylcholanthrene (MCA)-induced sarcoma,
MCA 101, has a very poor ability to present endogenously
generated viral antigens (14). This tumor grows rapidly and
lethally in nonimmunosuppressed hosts (3, 8). Furthermore,
of seven MCA-induced tumors generated and characterized
extensively in our laboratory (3, 6, 7, 8, 14, 28), only MCA
101 can neither act as an immunogen in vivo nor generate
CTL in vitro (3, 8). In the current studies, we retrovirally
transduced these cells with the cDNA coding for mlFN-%
Our aim was not to use IFN-7 in these experiments for its
effects on the immune system, but instead, for its effects on
the antigen presentation capabilities of a tumor cell. We show
that MCA 101 can be converted from a poor presenter of
endogenously generated antigens to a tumor capable of
efficiently presenting antigens. Furthermore, this conversion
is correlated with a change in the tumor's ability to elicit
tumor inf~trating lymphocyte (TIL) cells. Whereas wild-type
tumor does not elicit TILLcells, tumor retrovirally transduced
with the cDNA for IFN-7 does. Most importantly, these
TIL are active against the low class I-expressing wild-type
MCA 101. These experiments represent the first reported to
use functional assays to study the antigen presentation capabilities of a tumor before and after gene modification, and
correlate this modification with in vivo behavior.
Materiah and Methods
Tumor and Animals. MCA-induced sarcomas (101and 102)(3)
were generated in our laboratory in 8-wk-old female C57BL/6n
(B6)mice(Animal Production Colonies, FrederickCancer Research
and Development Futility, National Institutes ofHealth, Frederick,
MD) by intramuscular injection of 0.1 ml of 0.1% 3-MCA in se-
sameseedoil. Tmnor lineswere passagedin 136mice, and alltumors
were of early transplantation passage (passage 3-7). Tumors were
harvestedfrom mice,triph-enzyme digestedwith 0.1% collagenase,
i Abbreviationsusedin this~t,er: CM, completemedium;ER.,endoplasmic
reticulum;LAK,lymphokine-activatedkillercells;MCA,methylcholanthrene; MCN, meanchannelnumber;Nco K, neomycinresistance;NP,
nudeoprotein;PKS,wild-typeinfluenzavirusA/PK/8/34; TATA,tumor-
associatedtransplantation antigen;TIL, tumor-infaltratinglymphocytes;
Vac,vaccinavirus; WT, wild-type.
0.002% DNase, and 0.01% hyaluronidase (all from Sigma Chemical Co., St. Louis, MO), and maintained in monolayer culture in
complete medium (CM) containing RPMI 1640, 10% heat-inactivated FCS, 0.1 mM nonessential amino acids, 1.0 mM sodium pyrnvate (all from Biofhids Inc., Roekville, MD), 5 x 10-s
M 2-ME (Aldrich Chemical Co., Milwaukee,WI), and 0.03% (100
raM) glutamine (National Institutes of Health Media Unit,
Bethesda, MD).
Retmuiral Transductionof TuraorLines. A long-term cultured line
of MCA 101, was cloned by limiting dilution techniques. All 41
clones tested expressed low levelsof class I. One of these clones
was designated 101.WT and was plated at 10~cellsin CM in a 75-cm2tissue culture flask (Falcon; Becton Dickinson Labware, Oxnard, CA) and allowed to adhere overnight. To create the bulk-transduced cell line 101.NAT, viral supernatants were added to
washed adherent 101.WT tumor cellsin the presence of polybrene
(8 #g/ml) for 12 h at 37~ as described previously (29). CM was
added the next morning. The retroviral construct used was obtained from E. Gilboa (Memorial Sloan-Kettering Cancer Center,
New York, NY) and hasbeendescribedpreviouslyunder the designation DC/TKIFN~ (30), but which was also referred to as
NAT/IFN-~/by the Gilboa group, and is simply called NAT in this
manuscript. Briefly, the construct was derived from the genome
of the Moloneymurine leukemia virus and contained the mlFN-~/
gene, which was under the transcriptional control of the tbymi-
dine kinase promoter, and the neomycin resistance geue (Neo IL)
gene, which was under the transcriptional control ofthe LTR from
the Moloney murine leukemia virus. The bulk-transduced cellline
101.NAT was cloned and two high class I-expressing clones
(101.22H and 101.28H), and two low class I-expressing clones
(101.18Land 101.25L)were selectedfor further work. A bulk control cell line transduced with the Neo K gene alone and called
101.LXSNwas generatedin the sameway,eacept that the retrovirus
used was designated LXSN. This virus is the Moloney murine
leukemia virus backbone with the Neo R gene alone, and was obtained from Dr. D. Miller (Fred Hutchinson Cancer Research
Center, Seattle, WA). Cells were selectedin 1 mg/ml (0.5 mg/ml
active)G-418 sulfate(Geneticin; Gibco Laboratories, Grand Island,
NY) for 2 wk and then cloned at 0.2 cells/well.Successfulintroduc-
tion of cDNA was verifiedby Southern blotting after selection of
the transduced cells in G-418. mlFN-~ activity in supernates of
transduced tumor cellswas measuredbyELISA(Quantlkine; R&D
Systems, Inc., Minneapolis, MN).
FACS~ Analysis. Cell surfaceclass I expression was measured
by FACS| analysis using a FACScan| 440 (Becton Dickinson &
Co., Mountain View,CA). Freshtumors wereharvested and triple-
enzyme digested vide supra before st~ining with appropriate antibodies for FACS| Cultured tumor cell lines were harvested with
0.02% ~ , washed, then stained for 30 rain with culture supernatant from the hybridoma 28.8.6s (anti-Kb and Db) (31), obtained from Dr. D. Sachs(MassachusettsGeneralHospital, Boston,
MA), or with the isotype-matched(IgG2a)control antibody (Becton
Dickinson & Co.) followed by goat anti-mouse FITC-conjugated
antibody (Boehringer Mannheim Biochemicals, Indianapolis, IN).
FACS| analyseswere standardized with Calibtite flow cytometer
beads (Becton Dickinson & Co.).
Generation of TIL and Anti-viral E.~ectorCells, TIL were generated from tumors that were injected subcutaneously and harvested
"~2wk later. They were enzymatically digested asdescribed above,
then washed twice with HBSS (Biofluids Inc.). Thy 1.2-bearing
cells within the tumor digest were separated by immunobeading
using magnetic beads coated with Thy 1.2 antibody and cultured
in 20 U/ml of rhIL-2 (Cetus Corp., Eme~filh, CA) as previously
1424 Enhancementof Antigen Presentationof NonimmunogenicTumor
described(7, 28). AntiviralpolyelonalresponderT cellpopulations
were generated as describedpreviously(14). Briefly,female6-wk-old B6 micewere primed with 20 hemagglutinin U of PR8 viral
allantoicfluid.Afteratleast2 wk, spleenswereremoved,dispersed,
and cultured in IMDM with 7.5% heat-inactivatedFCS(Biofluids
Inc.). Splenocyteswerethen cultured for 7d in the absenceofIL-2,
either with syntheticpeptideat 1#g/ml ('~0.6 #M), or with PR8-infected spleen cells (14).
In Vivo TumorTherapyExperiments. 8-12-wk-oldC57BL/6n
mice were irradiated with 500 tad and then injectedintravenously
with 3 x 106 wild-type MCA 101, 101.22H, or 101.28H (high
class I-expressing clones of the mlFN-3, gene-modified cell line
101.NAT) tumor cells in 1 ml of HRSS to induce pulmonary
metastases. On day 3, TIL were injected at doses specifiedin the
text. Where six'dried, mice were treated intraperitoneally with
10,000 U rlL-2 in 0.5 ml HBSS twice daily for fiveconsecutive
days.On day 14, after tumor injection, all mice were ear tagged,
randomized, and killed. Pulmonary metastaseswere enumerated
in a coded, blinded fashion as describedpreviously(3). If pulmonary metastasese~eeded 250, theywere deemedtoo numerous to
count (TNTC).
VirusesandInfectionof TumorLines. 3 x 106tumor cellswere
washed three times in RPMI 1640, then placedin 1 ml of RPMI
with 0.1% BSA(SigmaChemicalCo.), and30mM Hepes(Biofluids
Inc.) at pH 6.8. Cells were then infectedfor 90 min at 37~ with
20 PFU/cell of recombinant vacciniavirus constructs containing
either the influenza A/PR/8/34 nucleoprotein (NP) gene, or a
non-cross-reactive control gene, neuraminidase (NA) (32). After
infection, target cellswereincubated in CM for 3 h at 37~ then
labeledwith SlCras describedbelow. The production of the viral
antigens, as a measure of infection by vaccinia,was quantified by
FACS| with the antivaccinia mAb designated TWII, kindly
providedby Drs. J. YewdellandJ. Bennink (National Institute of
Allergy and InfectiousDiseases,Bethesda,MD), followedby goat
anti-mouseFITC-labeledantibody (GAMF)(BectonDickinson &
Co.). Cell membranes were rendered permeable by fixation with
0.5% paraformaldebydefor 20 rain at room temperature, stained
with TWII mAb in the presenceof 0.1% saponin (SigmaChemicalCo.) in PBS,then stainedwith FITC-labeledgoat anti-mouse
antibody as a second reagent as described previously (14).
51 51
Cr ReleaseAssays and CytokineReleaseAssays. 4 h- Cr release assayswere performed as describedpreviously(6). Briefly,2
x 106tumor targets in 0.5 ml of CM were labded with 200 #Ci
of SlCr (New England Nuclear, Boston, MA) for 90 min. For
studies of endogenous presentation, cellswere infectedwith virus
before SlCr labeling as detailed above. Labeledtumor cells were
coincubated with antiviral effectorCTL lines or TIL for 4 h. Su-
peruatantswereharvestedandcountedwith agammacounter (LKB
Instruments, Inc., Gaithersburg,ME)).Percentlysiswas calculated
asfollows:100 x [(experimentalcpm - spontaneouscpm)/(maximal cpm - spontaneous cpm)]. Effector recognition of tumor
targets by measuring cytokine releasewas done using tumors as
described pz~iously (7). Briefly, 5 x 10s effector ceils were incubated with 106 stimulator ceils for a period of 18-24 h in 24-well tissue culture plates. Supernatants were then harvested and
assayedfor cytokine production.
Results
TransfectionofMCA 10I with mlFN-T. We have recently
demonstrated that a routine MCA-induced sarcoma, MCA
101, was a poor presenter of endogenously generated viral
antigens in the context of MHC class I molecules to CD8 +
CTL. This defect was found to be reversed by incubation
with exogenous mIFN-7 (14). The question remained as to
whether MCA 101 was nonimmunogenic because it lacked
tumor-associated transplantation antigens (TSTA), or because
of its failure to present somepostulated, but asyet unidentified,
TSTA.
Since the amelioration of MCA 101's ability to present endogenous antigens after treatment in vitro with exogenous
mIFN-3, was transient, and correlated with the duration of
class I expression (data not shown), we hypothesized that
retroviral transduction of MCA 101 with the cDNA for
mIFN-3, would have a prolonged effect on the processing
and presentation of endogenous antigen by virtue of the stable
insertion of the gene. Thus, a Moloney leukemia retroviral
construct, designatedNAT, containingboth the mIFNw gene
and the Neo R gene was used. Bulk MCA 101 was cloned
by limiting dilution to generate 41 different clones, all of which
expressed low quantities (mean channel number [MCN] <10)
of surface MHC class I molecules. We then retroviraUy transduced one of these wild-type dones, designated 101.WT, with
the cDNA for mlFN-3,, and selected it in G-418 for 2 wk
to create the cell line designated 101.NAT. In parallel, 101.WT
was transduced with a Moloney leukemia virus vector, designated LXSN, containing the Neo R gene without the IFN-3,
gene, to create the cell line 101.LXSN, which was treated
in exactly the same way in all respects as 101.NAT. Although
the bulk-transduced tumor line 101.NAT secreted levels of
mlFN-3, that were consistently <5 U/106 cells per 24 h, as
measured by ELISA, it did express much higher levels of surface class I molecules as measured by FACS| analysis than
the Neo R bulk-transduced cells 101.LXSN and 101.WT (Fig.
1). This pattern of verylow or unmeasurable mlFN-3, production with greatly increased classI production is not dissimilar
to that seen by Gansbacher et al. (30). As shown in Fig. 1,
increased class I expression was seen in both fresh and cultured tumor lines. 101.WT and 101.LXSN showed two populations of cells in fresh preparations. These data were interpreted to indicate that the lower class I expressing cells in
these preparations were tumor cells and that the higher class-I
expressing cellswere likely to be infdtratingT cellsand stromal
cells. Extensive immunohistochemicalstudies of subcutaneous
nodules, as well as pulmonary metastases of MCA 101, have
been done using the same antibody as was used for FACS|
analysis (28.8.6s). In these studies, MCA 101 tumor cells were
found to retain their low expression of class I molecules in
vivo (Dr. JeffreyS. Weber, National Cancer Institute, Bethesda,
MD, personal communication). Expression of the Neo R gene
was evidenced by selection of the transduced cell line in 1
mg/ml of G418. Tumor cells that did not receive the Neo
R gene could not be grown in G418 (data not shown). Integration of the Neo R gene was shown by Southern transfer
at 3 mos after transduction (not shown).
Presentationof ViralAntigens bymIFN-y Gene-modifiedMCA
I01 toAntiviral CTL. In an effort to test the effect of IFN-'y
gene transduction on the ability of MCA 101 to present endogenous antigens, we infected IOI.WT, IO1.LXSN, and
1425 Restifo
et al.
Cultured Fresh
400
0 ....~ . --~ . --~ 0
104 10~ 102 103 104
o ~....x~...................
104 lo ~ loz lo~ lo 4 104 lo ~ lo2 104 lo4
104 101 102 103 104 104 101 102 104 104
CHANNEL FLUORESCENCE
Figure 1. FACS
~ profilesofthefollowingthreebulktumorcelllines:
101.WT(wild-typeMCA 101),101.LXSN(MCA101 transducedwith
the Neo R genealone),whichwas madefroma done of 101.WT,and
101.NAT(MCA101transducedwith theNeoK geneandwith IFN-'y),
whichwasalsomadefromthesamecloneof101.WTas101.LXSN.Profiles:
(left)culturedtumorlines;(right)freshtumorpreparations,whichhave
beenharvestedfromsubcutaneoussitesinmiceandtripleenzyme-digested
asdescribedinMaterialsandMethods.Visiblein theprofilesofthefresh
preparationsis abimodalcurve(top).The high-expressingcellsin these
curvesprobablyrepresentstromalcellsalongwith infiltratinghostim-
munecells.
101.NAT with wild-typeinfluenzaA/PR/8/34 (PR 8), then
subjected them to killingby appropriately directed CTL. Gene-modified 101.NAT was killed more efficiently in this assay
(25% at an E/T of 30:1) than were 101.WT and 10LLXSN
(<5% at an E/T of 30:1). However, FACS| data showed poor
infection of 101.NAT compared with the other two cell lines
(data not shown), suggesting that the endogenous mlFN-'y
gene made these tumor cells less susceptible to infection by
the PR8 virus. This observation is consistent with the observed suppressive effect of exogenous IFN-3: on viral infection with influenza A (33).
To overcome this problem, we repeated our experiments
using vaccinia viruses genetically engineered to contain
influenza A genes. Effector cells were splenocytes from B6
mice stimulated in vivo with Vac-NP and in vitro with the
NP peptide (amino acids 365-380) (34, 35). As shown in
Fig. 2, wild-type MCA 101 (101.WT) was not killed when
it was sham infected (no virus, NV), when it was infected
with the control vaccinia virus containing the NA gene, or
when it was infected with the vaccinia virus engineered to
express the influenza A NP gene. IFN-3/gene-modifiedtumor
cells (101.NAT) were also not killedwhen sham infected (NV)
or when infected with a control virus (Vac-NA), but were
specifically killed when they were infected with the Vac-NP
virus. Like 101.WT, 101.LXSN (not shown) did not present
viral antigens, indicating that the improvement in 101.NAT's
ability to present antigen was not due simply to the effects
100
8O
0
4O
ae
101.WT
NV NA
VAC
10Sl lOS1
NP NV
VAC
1031 1031
E:T Ratio
101.NAT
V~
1031 1031
Figure 2. s~Crreleaseassayinwhichwild-tyl~101.WT,(left),or IFN-3'
gene-modified101.NAT,(righ0, aresubjectedto killingbyCTLdirected
againsttheinfluenzaANP geneproduct.Cellswereeithershaminfected
(NV),infectedwithanengineeredvacciniaviruscontainingtheinfluenza
A hemagglutiningeneas a control(ContVat),or infectedwith an en-
gineeredvacciniaviruscontainingtheinfluenzaAnucleoproteingene(NP
l-tic).Onlythe mlFN-'),gene-modifiedMCA 101 tumorinfectedwith
the appropriateviruswaskilledin thisassay.Thisfindingdemonstrates
that the modificationof MCA 101with the genefor IFN-3'abrogates
its inabilityto presentendogenousantigen.
of retroviral transduction. It thus appeared that by inserting
the mlFN-3~ gene into a bulk population of MCA 101, we
could convert it from a poor presenter of antigen to a cell
line capable of presenting antigen to a similar extent as our
most immunogenic tumors (see reference 14). This finding
led us to hypothesize that MCA 101 might be capable of
presenting its own tumor antigens in vivo.
Generation of TIL from mlFN-'y Gene-modified MCA
101. Generation of successful cultures of TIL from wild-type MCA 101 has never been achieved in our laboratory.
To test whether the done being used in the present studies
was similar to the parental line, tumors were implanted sub-
cutaneouslyin syngeneic B6 mice, and 10-20 d later TIL were
harvested. Two of two blind attempts at the establishment
of TIL cultures from wild-type MCA 101 (101.WT) or
MCA 101 transduced with the Neo R gene alone (101.LXSN)
yidded no successfulTIL cultures. Although bulk-transduced
101.NAT tumor at first appeared to be successful, in two of
two blind attempts the TIL cultures were ultimately overgrown with tumor. We hypothesized that some low class
/-expressing tumor clonotypes, which may be the rare cells
visible in Fig. 1, existed in the bulk population, and that they
were able to overgrow the culture before TIL could be established.
To obtain pure cultures of uniformlyhigh classI-expressing
transductants, the bulk-transfected MCA 101 (101.NAT) was
cloned. We found that the insertion of the cDNA for IFN-'y
caused increased expression of steady-state surface class I as
measured by FACS~ analysison some, but not all, of the tumor
clones tested (Fig. 3). Two high (H) class1--expressingclones,
101.22H and 101.28H, and two low (L) class I-expressing
clones 101.18L and 101.25L, were selected from bulk-1426 Enhancement
of AntigenPresentationof NonimmunogenicTumor
1501~ 101.18L
(MCN:8)
ol
101.22H i
(MCN:883)
I I
1~ 100 10' 10' 1(~ 10'
IS0~ t0t,25L lS0
l 101.28H
) (MCN:
839)
o ]
100 10' 10' 10' 104 IO
~ I0' 1~ I~ 1
10'
Figure 3. Neitherthe bulk-transducedtumor,nor any of the clones
fromthebulk-transduced101.NATproducedmeasurablelevelsofIFN--y
byELLS&However,bnlk-transducedcellsusingtheIFNwvectorexFessed
muchhigherlevelsof surfaceclassI thanthosecallstransducedwith a
controlvectorexpressingonlythe Neo ILgene.FACS
| profihsof two
celllinesclonedfromthebulk-transducedMCA101tumorafterselection
inG418.WithaMCNof8,done 18Lwasamongthoseclones~pressing
the lowestlevelsof classI afterstainingwith the mAb 28.8.6s(which
stainsboth Kband Db). FITClabeledgoat anti-mouseAb was usedas
asecondreagent.C]one22H,on theotherhand,wasamongthehighest
classI expresserswithaMCNof882.Theexperimentshownwasrepeated
twicewithsimilarresultsoveraperiodofseveralweeks,showingstability
of the phenotype.
transduced 101.NAT. Despite their variable levels of class I
expression, none of these clones secreted >5 U of mIFN-3,/106 ceUs/ml per 24-h ELISA. These clones were then implanted subcutaneously into syngeneic B6 mice. There were
no significant differences between the growth rates of high-
and low-dass-I expressing tumors in vivo (data not shown).
When subcutaneous tumors were harvested after 10-20 d,
Thy 1.2+ cells were obtained by immunobeading of single
cell suspensions of tumor cells. In two of two blind attempts,
we could not grow TIL from low class I-expressing tumor
clones (101.18L or 101.25L) of the mIFN-'y-transduced
m
E
z
250
100
MCA 101.22H
i i i TNTC
, , ,
Sslifte I 2 106 3 06
2~
10(]
MCA 101.~
t |
T
Saline JL2- 106 3x106 107
Treatment
Figure 4. In this experiment,tumorwas injectedon day0 to create
lungmetastases.On day3, micereceivedoneofseveraltreatmentslisted.
On day 14, the numberof pulmonarymetastaseswereenumeratedand
areex~essed.CD8+TILL
generatedfromahighclassI-~=pressingmlFN-
-y-transducedtumorclone(101.221-1)
areeffectivenotonlyagainstthegene-
modifiedtumor,but againstthe wild-typetumor (101.WT)as well
101.NAT. On the other hand, in six of six attempts, we could
consistently grow TIL from high class I-expressing tumor
clones of mIFN-'y-modified 101.NAT.Like other routine TIL
generated in our laboratory, the TIL cultures used in vivo
in Fig. 4 were found to be exclusively CD8 + by FACS|
(data not shown). These antitumor CTL were then expanded
in vitro in 20 U/ml of rhII.,2 and tested in vivo and in vitro.
In VivoEffectivenessof TIL Generatedfrom ralFN-7 Gene-modifiedMCA 101. In an effort to test whether the TIL
generated from high classI-expressing mIFN-3,-transduced
clones could be active in an adoptive immunotherapy model
against established wild-type tumor in vivo, we used a 3-d
lung metastases model (3). In these experiments (Fig. 4), mice
were injected intravenously with fresh 101.WT or 101.22H
tumor ceils. On day 3, mice were treated with either saline
alone, IL-2 (10,000 U) in saline twice daily for 5 d, or the
same dosage of IL-2plus varying dosages of TIL. Mice were
kilhd on day 14 when their lungs were harvested and counted
in ablind fashion for the number ofpulmonarytumor nodules.
As shown in Fig. 4, TIL generated from the mlFN-3~ gene-modified 101.22H tumor were effective against pulmonary
metastases from the 101.22H tumor. Most significant, however, was the effectiveness of these CD8 + TIL against the
wild-type MCA 101 tumor. In the experiment shown, the
average number of pulmonary metastases from 101.WT were
reduced from 200 in the Ib2 alone group to 48, 8, and 0
at 106 TIL, 3 x 106, and 107 TIL, respectively. TIL generated from mlFN-~/gene-modified, high class I-expressing
tumor done 101.22H were therapeutically effective against
the unmodified tumor in three other repeats of the same experiment.
SpecificityofAnti-MCA I01 TIL In Vitr~ Four SlCr release assays were performed on the TIL generated from
101.22H tumor. Only two of four T1L cultures obtained from
mIFN-'y gene-modified tumor were specifically cytotoxic.
However, the two cultures that were cytotoxic demonstrated
relative tumor spedfidty (Fig. 5). Two TIL cultures that were
not lytic against fresh tumor preparations in vitro were still
therapeutic in vivo. This finding was consistent with previous
data (9) that showed that CD8 + TIL that were therapeutic
in vivo were not in every case cytotoxic.
TIL generated from IFN-'y-modified tumor were treated
for release of cytokines after coincubation with appropriate
tumor ceils. TIL derived from 101.22H tumor were tested
by exposure to a variety of stimuli. The results shown in Fig.
6 made use of the same cells as were used one day earlier
in the therapy experiment depicted in Fig. 4. Maximal release was gauged when TIL were stimulated with an anti-CD3 antibody, 2Cll. Background levelsof TNF-ot and IFN-'y
when TIL were incubated alone (without stimulation) were
relatively low for both cytokines (Fig. 6). Fresh tumor preparations alone were found to secrete <(5 U of IFN-3,/106
cells/m1 per 24 h, with the exception of 101.22H which produced 6.3 U IFN-'y/106 cdls/ml per 24 h. Production of
TNF was variable by fresh tumor preparations, but was uniformly <20 U/106 cells/m1 per 24 h. Values obtained from
tumor alone were subtracted from values obtained from tumor
1427 Restifo
et al.
_r
0
re
~3
r
80
60
40
20
101.WT 101.WT
+ Exog 7
't
\
\
\
101.2; 102,WT
.-.1...
",,._.
1031 1031 1031 1031
E:T Ratio
102.WT
+ Exog 7
1031
102.NAT
1031
101,18L
1031
Figure5. To determine whether the TIL generated from the mlFN-~/
gene-mMitied MCA 101 tumor (101.22II)were tumor specific,a 4-h 51Cr
releaseassaywas doneusinga variety oftargets includingwild-typeMCA
101 (101.WT), 101.WT after a 48-h pretreatment with 200 U/ml of
nnlFNw (101.WT + Exogenous3,),IFN-3,gene-modifiedMCA 101done
22H, which is a high expresser of class I (101.22H), or a control tumor,
MCA 102 (102.WT) treated exactly as was MCA 101. As an additional
control, we included a clone of MCA 101 (101.18L),which is IFN-3'gene
modified in the sameway as was done 101.22H, but which expresseslow
levelsof class I. 101.WT + Exogenous 3' and 101.22H were lysed relatively more dficiently than controls. This finding indicates that tumor-
associatedantigen isbeingrecognizedwhich isrehtively specificfor 101.WT,
and which requires relatively high class I expression for killing in a 4-h
slCr release assay.
plus TIL. The most important finding was that 101.WT
stimulated release of IFN-% as well as t01.22H, and release
ofTNF-o~was almost asgreat when stimulated with 101.WT
as with 101.22H. 102.WT, another MCA-induced sarcoma,
was prepared in exactly the same way and used as a control.
The primary purpose of this control was to seewhether the
cytokine rdease for the 101tumor was specific.
Control tumor
102.WT stimulated substantially less cytokine release from
these TIL than did 101.WT or 101.22H. These results indi-
UNITSIml TNF (~)
STIMULATOR 100 200 300 400
. . . . ;;~
2C11 mAb ~u~ ./~
NONE
101.22H I
101.WT I
102 WT m Z~]
i i I I (~0//
20 40 60 80 1
UNITS/ml IFN-y (~)
Figure 6. TILLgenerated from 101.22H were tested for the capacity to
releaseTNF and IFN~ to the tumor stimulators listed. 2Cll mAb (anti-CD3) was used to test maximal release,and CTL alone (None)was used
to test spontaneous release.This graph showsthat 101.WTstimulated TIL
to releaser (mlFN-qr and mTNF-~) nearlyaswell as its mIFN-3,
gene-modif~counterpart 101.22H.Asin the caseof the cytotoxicityassay
depicted in Fig. 5, these TIL were found to be relatively, but not abso-
lutdy specific(note some releaseofcytokines against the 102.WT tumor).
cared significant specificityin the response and furthermore
suggested a possible mechanism for the action of the TIL
cellsin vivo.These experiments were repeatedwith high class
I-expressing, transduced tumor clone 101.28H with similar
results. Thus, TIL generated from mIFN-% gene-modified
tumor clones could be triggered to secrete, in a relatively
tumor-specific fashion, IFN-qr and TNF-oe against not only
high class I-expressing tumor clones, but also against the
low class I expressing-wild-type tumor.
Discussion
A major problem with cellularimmunotherapies ofcancer
in humans hasbeen the limitedability to generate therapeutic
cells with antitumor properties in vivo. A recent approach
to this problem has been to exploit the immunomodulatory
effectsof genetic modification of tumor cells with cytokine
genes, including the IFN-3, gene (30, 36-38). The MCA 101
murine sarcoma maybe a model for many tumor histologies
because it has no demonstrable immunogenicity(3, 14). We
show herethat retroviral-mediated transduction of wild-type
MCA 101 with mlFN-~/cDNA enables the generation of
CD8 + TIL, which cannot be generated from the wild-type
tumor. Most importantly, theseTIL were therapeuticallyactive
in vivo against both the gene-modified and the wild-type
tumor in a 3-d lung metastasesmodel. Although the 101.WT
tumor was not consistently lysed by therapeutic TIL, it did
stimulate the releaseofcytokines with relative specificity.Finally,mlFN-3, gene-modifiedMCA 101tumor presentedviral
antigens much more efficiently than controls, suggesting a
mechanism for these observations.
Although mlFN-qr transduced-tumorlines and clones secrete little, if any,mlFN-3, into culture supematant, the effect
of the cytokine could act primarily via intracdlular IFN-3'
receptors. Basedon our experiments that show steady state
increases in the expression of cell surface-dass I molecules,
and on the enhanced killing by CD8 + T cellsin experiments
using a defined antigenic system, we may infer that there is
increased density of tumor antigen peptide/class I molecule
complexes on the cell surface.This inference is strengthened
by recent evidence which shows that antigen presentation
maybe enhancedat severalimportant points by the mlFN-'y
molecule. In addition to increasing classI H chain at the level
of transcription initiation (39), IFN-3, has alsobeen shown
to increase steady state levels of TAP 1 and 2, the putative
peptide transport proteins, aswell asthe geneproducts similar
to the low-molecular masspolypeptide (LMP) or proteasome,
which maybe involved in cleavingproteins into peptides be-
forefeedingthem into the EK or post-EK compartment where
they bind to class I molecules (27). Finally, IFN-3, has been
shown to be a potent upregulator of certain accessorymole-
cules, ICAM-1 among them (40), and while not specifically
addressed in the studies presented here, the effect of IFN-3,
on these molecules may play an important role in the phenomena described here. The relative contributions of these
mechanisms are currentlybeing addressed in our laboratory.
Increasedantigen densityon mlFN-~, gene-modifiedMCA
101 may explain why a therapeutically useful CD8 + T cell
1428 Enhancement of Antigen Presentation of NonimmunogenicTumor
response can be obtained from high class I-expressing clones,
and not from low class I-expressing mlFN-3( transduced
clones, control Neo R-transduced clones, or from wild-type
tumor. There is evidence in tumor immunology that an immunologically strong afferent tumor stimulus can elicit a response against an immunologically much weaker efferent
stimulus (41). Alternatively, IFN-'y could act primarily
through upregulation of class II expression on macrophages
and dendritic cells, thereby enhancing antigen presentation
to Th cells (42).
During tumorigenesis, tumors that poorly present endogenous antigens might enjoy a selectiveadvantage since mutations of normal cellproteins and expression of oncogeneprod-
ucts may yield antigenic peptides that could result in the lysis
of the incipient cancer. It has recentlybeen demonstrated that
viruses can persist in neurons because they lack class I expression (43). However, the related concept that tumor cells can
escape immunosurveillance by T cells of tumor cells is controversial, and the role of class I molecules in this process
is unclear (44). In fact, there is evidence that absence of class
I makes some tumors more susceptible to lysis by NK cells
(45). Nevertheless, MCA 101 is an extremely virulent tumor
that may have benefited during its genesis from its poor
antigen-presenting abilities (14). The explanation of MCA
101's poor capability to present endogenous antigen, which
appears to be entirely reversible with IFN-% may stem from
its possession of a gene product not unlike the E1A protein
of the adenovirus (46), as has been suggested for small cell
lung cancer (47). It is in fact known from other studies, that
MCA 101 suppresses class I production at the level of transcription initiation (48).
Our findings may have strategic implications for the development of immunotherapies for cancer in humans. A candidate for this approach may be a human tumor that fails
to present endogenously synthesized antigen because it has
decreasedclass I synthesis and normal class I transport, which
can be upregulated by treatment with hlFN-% This tumor
could then be used to establish a line in vitro that would
then be transduced with the gene for IFN-3', subsequently
reimplanted at a subcutaneous site, and then harvested several weeks later for TIL. Thus far, our studies in this direction (N. P. Restifo, unpublished observations) suggest that
some human tumor histologies, such as small celllung cancers,
may have antigen presentation defects very similar to those
described previously for MCA 101: specifically, those with
low steady state expression of dass I, but which have normal
transport (14). Thus, our work with the murine tumor MCA
101 suggests that human tumor cells previously considered
to be nonimmunogenic could be rendered better presenters
of endogenous antigen by IFN-3~ gene modification. This
approach may be useful in generating CD8 + antitumor
effector T cells against tumor histologies not previously
thought to be susceptible to T cell-based immunotherapies
of cancer.
The authors wish to thank Jonathan Yewdell,Jack Bennink, Eli Gilboa, Stuart Marcus, Barbara Pockaj,
Patrick Hwu, and Robert Cowherd for their excellent help and stimulating discussion.
Addresscorrespondence to Nicholas P. Restifo, Surgery Branch, Division of Cancer Treatment, National
Cancer Institute, National Institutes of Health, Bethesda, ME) 20892.
Received for publication I2 December I991 and in revised.form 11 February 1992.
~ferences
1. van der Bruggen, P., C. Traversari,P. Chomez, C. Lurquin,
E. De Plaen, B.Vanden Eynde, A. Knuth, andT. Boon. 1991.
A gene encoding an antigen recognized by cytolytic T lym-
phocyteson ahumanmelanoma. Science(Wash.DC). 254:1643.
2. Itoh, K., C.D. Platsoucas,and C.M. Balch. 1988.Autolognus
tumor-specific cytotoxic T lymphocytes in the infiltrate of
human metastatic melanomas: activationby interleukin-2 and
autologous tumor cells,andinvolvementofthe T cellreceptor.
J. Exi~ Med. 168:1419.
3. Mul6,J.J., J.C. Yang, R. Lafreniere,S. Shu, and S.A. Rosenberg. 1987. Identification of cellular mechanisms operational
in vivoduring the regression of establishedpulmonarymeta-
stasesby the systemicadministration ofhigh-doserecombinant
interleukin 2. J. Immunol. 139:285.
4. Greenberg, P.D., M.A. Cheerer, and A. Fefer. 1981. H-2 restriction of adoptive immunotherapy of advanced tumors. J.
Immunol. 126:2100.
1429 Kestifo et al.
5. Schreiber,H., P.L.Ward, D.A. Rowley,andHJ. Stauss.1988.
Unique tumor-specific antigens. Annu. Rev. Iramunol.6:465.
6. Barth, R.J., S.N. Bock, J.J. Mul~,and S.A. Rosenberg. 1990.
Unique murinetumor-associatedantigens identifiedby tumor
infiltrating lymphocytes.J. Imraunol. 144:1531.
7. Barth, R..J., Jr., J.J. Mul~, p.J. Spiess, and S.A. Rosenberg.
1991. Interferon 3' and tumor necrosis factor have a role in
tumor regressionsmediatedby murine CD8 tumor-infiltrating lymphocytes,f. ExF Med. 173:647.
8. Weber,J.S., G.Jay,K. Tanaka,andS.A.Rosenberg. 1987.Im-
munotherapyofaroutinetumor with interleukin-2: increased
sensitivity after MHC class I gene transfection. J. ExF Med.
166:1716.
9. Tolyalhn,S.L.,D. Solomon,andS.A.Rosenberg. 1989.Tumor-
specificcytolysisbylymphocytesinfiltratinghuman melanomas.
f. Irarnunol. 142:3714.
10. Rosenberg, S.A., B.S.Packard,P.M. Aebersold,D. Dolomon,
S.L.Topalian,S.T.Toy,P. Simon, M.T.Lotze,J.C. "fang,C.A.
Seipp, et al. 1988.Use of tumor-infiltrating lymphocytesand
interleukin-2in theimmuno~erapyofpatientswith metastatic
melanoma. N. Engl. J. Med. 319:1676.
11. Kosenberg, S.A., P. Spiess, and K. Lafreniere.1986. A new
approachto the adoptiveimmunothenpy ofcancerwith tumor-infiltrating lymphocytes. Science(Wash.DC). 233:1318.
12. Greenberg,P.D.1991.AdoptiveT celltherapyoftumors: mech-
anismsoperativein the recognition and elimination of tumor
ceils. Adg lmmunol. 49:281.
13. Hewitt, H.B., E.K. Blake,and E.S. Walder. 1976. A critique
of the evidencefor activehost defenseagainstcancerbasedon
personal studies of 27 murine tumors of spontaneous origin.
Br. f Canc~ 33:241.
14. Kestifo,N.P., E Esquivel,A.L. Asher, H. St6tter, K.J. Barth,
J.K. Bennink, J.J. Muir, J.W. YewdeU,and S.A. Kosenberg.
1991.Defectivepresentationofendogenous antigensbya mu-
fine sarcoma:implicationsfor the failureofan anti-tumor immune response,f Immanol. 147:1453.
15. &hlen, C., J. Bastin, H.G. Ljunggren, L. Foster,E. Wolpert,
G. Klein, A.K.M. Townsend,and K. I~rre. 1990.Resistance
to H=2-restrictedbut not to allo=H2-specificgraft and cyto-
toxicT lymphocyteresponsesinlymphomamutant.J. Immunol.
145:52.
16. Madden, D.K.,J.C. Gorga,J.L. Strominger, and D.C. Wiley.
1991.The structure ofHLA-B27revealsnonamer self-peptides
bound in an extended conformation. Nature (Lond.). 353:321.
17. T.S.Jardetzky, W.S.Lane,K.A. Robinson, D.K. Madden,and
D.C. Wiley. 1991. Identification on self peptides bound to
purified HLA-B27. Nature (Lond.). 353:326.
18. Van Bleek, G.M., and S.G.Nathenson. 1990. Isolation of an
endogenouslyprocessedimmunodominant viralpeptide from
the class I H-2Kb molecule. Nature (Long). 348:213.
19. Krtzschke, O., K. Falk, K. Deres, H. Schild, M. Norda, J.
Metzger, G. Jung, and H. Kammensce. 1990. Isolation and
analysisof naturally processedviral peptides as recognizedby
cytotoxic T cells. Nature (Lond.). 348:252.
20. Falk,K., O. Krtzschke, K. Deres,J. Metzger, G. Jung, and
H.-G. Kammensee.1991.Identificationofnaturally processed
viral nonapeptides allowstheir quantification in infectedcells
and suggests an allele-specificT cellepitope forecast.J. ExF
ivied. 174:425.
21. Brown, M.C., J. Driscoll, and J.J. Monaco. 1991.Structural
and serologicalsimilarityofMHC-linkedLMPandproteasome
(multicatalyticprotcinase)complexes.Nature (Lond.). 353:355.
22. Glynne, K., S.H. Powis, S. Bech, A. Kelly,L.-A. Kerr, and
J. Trowsdale.1991.A proteasomg~rdatedgentbetweenthe two
ABC transporterlociin the classII regionofthe human MHC.
Nature (Lond.). 353:357.
23. Deverson,E.V.,I. K. Cow, W.J. Coadwdl,J.J. Monaco,G.W.
Butcher,andJ.C. Howard. 1990.MHC classII regionencoding
proteins related to the multidrug resistancefamily of transmembrane transporters. Nature (Lond.). 348:738.
24. Monaco,J.J., S. Cho, and M. Attaya. 1990.Transport protein
genes in the murine MHC: Possibleimplications for antigen
processing. Nature (Lond.). 250:1723.
25. Spies,T., M. Bresnahan, S. Bahram, D. Arnold, G. Blanck,
E. Mellins, D. Pious, and K. DeMurs. A gene in the human
major histocompatibility complex classII region controUing
the classI antigenpresentationpathway.Nature(Lond.).348:744.
26. Trowsdale,J., I. Hanson, I. Mockridge, S. Beck, A. Townsend, and A. Kelly.Sequencesencoded in the classII region
of the MHC relatedto the 'ABC' superfamilyof transporters.
Nature (Lond.). 348:741.
27. Kobertson, M. 1991.Antigen processing:proteasomesin the
pathway. Nature (Lond.). 353:300.
28. Yang,J.C., D. Perry-Lalley,and S.A.Kosenberg.1990.An improved method for growing murine tumor=infiltrating lym-
phocyteswith in vivoantitumor activity.fBioLResponseModif.
9:149.
29. Asher,A.L.,J.J. Mu16,A. Kasid,N.P.R~stifo,J.C. Salo,C,M.
Keichert, G. Jaffe, B. Fendly, M. Kriegler, and S.A. Posen-
berg. 1991.Murine tumor cellstransducedwith the genefrom
tumor necrosisfactor-o~:evidenceforparacrineimmune effects
of tumor necrosisfactoragainsttumors.f Immunol. 146:3227.
30. Gansbacher,I~, K. Bannerji, B. Daniels, K. Zier, K. Cronin,
and E. Gilboa. 1990. Ketroviralvector-mediated-/-interferon
genetransferinto tumor cellsgeneratespotent and long lasting
antitumor immunity. Cancer Res. 50:7820.
31. Ozato, K., and D.H. Sachs. 1981. Monoclonal antibodies to
mouse MHC antigens. III. Hybridoma antibodiesreacting to
antigens of the H=2bhaplotype revealgenetic control of isotype expression,f lmmunol. 126:317.
32. Smith, G.L., J.S. Levin, P. Palese, and B. Moss. 1987. Synthesis and cellular location of the ten influenza polypeptides
individuallyexpressedbyrecombinantvacciniaviruses.Virology.
160:336.
33. Baglioni,C., and T.W.Nilsen. 1983.Mechanismsof antiviral
action of interferon. Interferon. 5:23.
34. Townsend, A., J. Kothbard, EM. Gotch, G. Bahadur, D.
Wraith, and A.J. McMichael.1986.The epitopesofinfhenza
nucleoprotein recognizedby cytotoxic T lymphocytescan be
defined with short synthetic peptides. Cell. 44:959.
35. Could, K.,J. Cossins,J. Bastin,G.G.Brownlee,and A. Townsend. 1989. A 15 amino acid fragment of influenza nucleoprotein synthesized in the cytoplasm is presented to class
I-restricted cytotoxic T lymphocytes.J. Exp. Med. 170:1051.
36. Kosenberg, S.A. 1991. Immunotherapy and gene therapy of
cancer. Cancer Res. 51:5074s.
37. Fearon, E.K., D.M. PardoU, T. Itaya, P. Golumbek, H.I.
Levitsky,J.W. Simons, H. Karasuyama,B. Vogelstein,and P.
Frost. 1990. Interleukin-2 production by tumor cellsbypasses
T helperfunction in the generation of an antitumor response.
Cell. 60:397.
38. Wantanabe, Y., K. Kuribayashi,S. Miyatake, K. Nishihara,
E.-I. Nakayama,T.Taniyama,and T.-A. Sakata.1989.Exogenous expressionof mouse interferon 3, cDNA in mouse neuroblastoma C1300 cellsresults in reduced tumorigenicity by
augmented anti-tumor immunity. Proc.Natl. A_cad.Sci. USA.
86:9456.
39. Sugita, K., J. L. Miyazaki, E. Appella, and K. Ozato. 1987.
Interferonincreasestranscriptionofa majorhistocompatibility
class I gene via a 5' interferon consensus sequence.Mol. Cell
Biol. 7:2625.
40. Dustin, M.L., R. tLothlein,A.K. Bhan, C.A. Dinarello, and
T.A. Springer. 1986. A natural adherence molecule (ICAM-1): induction by IL-1and interferon-'/, tissuedistribution, biochemistry and function. J. Imraunol. 137:245.
41. LeGrue, S.J., H.N. Ananthaswamy, and W.J. Simcik. 1989.
Afferent and efferent specificityin the induction and elicitation ofparentalcross-protectiveimmunitybyanimmunogenic
murine tumor variant: associativerecognition of a unique
tumor-specific antigen on somatic cell hybrids. Cancer Res.
49:4747.
42. Weaver,C.T., C.M. Hawrylowicz, and E.K. Unanue. 1988.
T helpercellsubsetsrequirethe expressionofdistinctcostimula-1430 Enhancement
of AntigenPresentationof NonimmunogenicTumor
tory signalsby antigen-presenting cells.Proc.Natl. Acad. Sci.
USA. 85:8181.
43. Joly, E., L. Mucke, and M.B.A. Oldstone. 1991.Viralpersis-
tencein neurons explainedbylackofmajorhistocompatibility
class I expression. Science(Wash. DC). 253:1283.
44. Elliot, B.E., D.A. Carlow, A. Kodricks, and A. Wade. 1989.
Perspectiveon the roleofMHC antigensin normal and malignant cell development. Adv. CancerRes. 53:181.
45. I~rre, K., H. Ljunggren, G. Piontek, and R.. Kiessling.1986.
SelectiverejectionofH-2-deficientlymphomavariantssuggests
alternative immune defensestrategy. Nature (Lond.). 319:675.
46. Kalvakolanu,D.V.R., S.K.Bandyopadhyay,M.L. Hatter, and
G.C. Sen. 1991.Inhibition ofinterferon-induciblegeneexpres-
sionby adenovirnsEIA proteins:block in transcriptionalcom-
plex formation. Proc Natl. Acad. Sci. USA. 88:7459.
47. Doyle,A., W.J. Martin, K. Funa, A. Gazdar,D. Carney,S.E.
Martin, I. Linnoila, F. Cuttitta, J. Mulshine, P. Bunn, and
J. Minna. 1985. Markedly decreased expression of class I
histocompatibility antigens, protein, and mKNA in human
small-cell lung cancer.J. Exlx Med. 161:1135.
48. Lassam, N., and G. Jay. 1989. Suppressionof MHC class I
RNA in highly oncogeniccellsoccursat the leveloftranscrip-
tion initiation. J. Immunol. 143:3792.
1431 Restifo
et al.

