Proc. Nati. Acad. Sci. USAVol. 89, pp. 6020-6024, July 1992Immunology

Detection of rare antigen-presenting cells by the lacZ T-cellactivation assay suggests an expression cloning strategyfor T-cell antigens

(antigen presentation/p-glactosidasc/major histocmpatibifity complex/ovalbumin)

JAANA KARTTUNEN, SARAH SANDERSON, AND NILABH SHASTRI*Department of Molecular and Cell Biology, University of California, Berkeley, CA 94720

Communicated by Harden M. McConnell, March 2, 1992

ABSTRACT The a/fl T-cell receptor recognies a com-plex ligand formed by the association ofantigenic peptides withmolecules ofthe major histocompatibility complex (MHC). Theinherent limitations of the conventional T-cell activation assaysused to detect these peptide/MHC ligands have, until now,hampered the development of expression cloning systems forT-cell antigens. To overcome these limitations, we have re-cently introduced a method for detecting ligand-induced acti-vation of individual T cells. This assay, which makes use of alacZ reporter construct, differs from conventional ligad-induced activation assays in that it allows the detection ofsingle, activated T cells in large pools of resting cells. Weapplied the IacZ assay to the problem of screening expressionlibraries, which requires the ability to detect ligand-bearingantigen-presenting cells when they are present at very lowfrequency. We show here that ligand-expressing antigen-presenting cells can be detected at frequencies of 1:103-104, alevel of sensitivity compatible with the screening of cDNAlibraries. Furthermore, by using as antigen-presenting cellsCOS-7 cells stably transfected with the murine Kb class I MHCmolecule, we demonstrate that transiently expressed ovalbu-min is efficiently processed and presented to an ovalbumin/Kb_specific T-cell hybridoma. IacZ expression is induced in adetectable number of cocultured T cells, even when the oval-bumin cDNA consists of only 1:104 of the total DNA used totransfect the COS cells. These results suggest that unknownT-cell antigens may be identified by screening cDNA librariesin MHC-expressing COS cells using lacZ-inducible T cells asindicators of peptide antigen expression.

The a/X3 T-cell receptor (TCR) recognizes small antigenicpeptides bound to major histocompatibility complex (MHC)molecules on the surface of antigen-presenting cells (APCs).For T cells of unknown specificity, especially those thatcause tissue destruction in autoimmune diseases (1), identi-fying the antigenic peptides remains a difficult undertaking.The primary reason for this difficulty lies in the inherent lackof sensitivity of the bulk T-cell activation assays used todetect peptide/MHC-expressing APCs. Because these as-says measure only the average state of activation within aT-cell population, they cannot be used to detect activated Tcells when these constitute only a small percentage of the Tcells being assayed.To overcome this obstacle, we have recently developed a

method, the lacZ assay, for detecting the ligand-inducedactivation of single T cells (2). The lacZ assay utilizes areporter construct consisting of the bacterial (3-galactosidasegene (IacZ) under the transcriptional control of the nuclearfactor of activated T cells (NF-AT) element of the humaninterleukin 2 (IL-2) enhancer (3, 4). When this reporter

construct is introduced into T-cell hybridomas, occupancy ofthe TCR triggers not only the secretion of IL-2 but also theproduction and intracellular accumulation of lacZ (2). lacZexpression can be visualized by loading the T cells with eitherfluorogenic or chromogenic substrates (5, 6). By using thelacZ assay, individual ligand-activated T cells can be readilydetected, even when they comprise a very small fraction ofthe total T-cell population.An important advantage of being able to detect small

numbers ofactivated T cells is that it also allows the detectionof correspondingly rare ligand-expressing APCs. This prop-erty of the lacZ assay suggested that lacZ-inducible T cellscould be used to develop an expression cloning strategy forT-cell ligands. For such a strategy to be feasible, the lacZassay had to be sensitive enough to detect antigen-expressingAPCs at the low frequencies expected in the primary screenof an expression library.To test whether the lacZ assay could be used for expres-

sion cloning of T-cell ligands, we developed a model systemconsisting of a lacZ-inducible T-cell hybrid specific for theovalbumin (OVA)/Kb ligand and a COS cell line stablytransfected with the Kb murine class I MHC molecule (Kb-COS). Transient expression oftheOVA protein was obtainedby transfecting the Kb-COS cells with an expression vectorencoding the OVA cDNA. The presence of the OVA/Kbcomplex on the surface of the COS cells was measured bycoculturing the transfected Kb-COS cells with the T-cellhybrid and assaying for lacZ expression. By diluting OVA-encoding cDNA with irrelevant DNA, we show that thelacZ/COS cell assay provides an exquisitely sensitivemethod for detecting antigen-encoding cDNA. Based on thedata presented here, we suggest that this system could beused to screen cDNA libraries for genes encoding T-cellantigens.

MATERIAL AND METHODSCell Culture. All cells were maintained in RPMI 1640

medium supplemented with 10% fetal bovine serum (IrvineScientific), 2 mM glutamine, 1 mM pyruvate, 50 ,.M 2-mer-captoethanol, penicillin (200 units/ml), and streptomycin(200 pug/ml) at 370C in a 5% C02/95% air atmosphere.

Cell Lines. To facilitate the generation of lacZ-inducibleT-cell hybrids, a lacZ-inducible derivative of BW5147 (Z.8)was made by transfecting the BW5147 cell line with theNFAT-lacZ reporter construct (3) using the methodologydescribed previously (2). The B3xZ.8 (B3Z) cell line was

Abbreviations: APC, antigen-presenting cell; FDG, fluorescein di-f3-D-galactopyranoside; HEL, hen egg lysozyme; IL-2, interleukin 2;MHC, major histocompatibility complex; NF-AT, nuclear factor ofactivated T cells; OVA, ovalbumin; TCR, T-cell receptor complex;X-Gal, 5-bromo-4-chloro-3-indolyl 3-D-galactopyranoside; FACS,fluorescence-activated cell sorting; SV40, simian virus 40.*To whom reprint requests should be addressed.

6020

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

0

Proc. Natl. Acad. Sci. USA 89 (1992) 6021

generated by fusing Z.8 with B3, a Vs5-expressing cytotoxicT-cell clone specific for the OVA/Kb ligand (7). The resultinghybridomas were screened for expression of the B3-derivedTCR by staining with a Vp5-specific antibody (see below). Torestore CD8 expression, which was extinguished in the T-cellhybrids, the V,5-expressing clones were transfected with aretroviral vector, kindly provided by D. Littman (Universityof California, San Francisco), containing the CD8a gene (8).The V135+ CD8+ hybridomas were screened for OVA/Kbspecificity in the 1L-2 and lacZ assays described below.The K89-7.OVA cell line was generated, as described

elsewhere, by transfecting a Kb-expressing subclone of theK89-7 L-cell line with the OVA cDNA (N.S., unpublisheddata).The Kb-COS cells were obtained by stably transfecting the

COS-7 cell line with a genomic clone of the murine Kbmolecule (I 497) (9), which was kindly provided by R. A.Flavell (Yale University School of Medicine). By using thecalcium phosphate method (10), COS-7 cells were transfectedwith a total of 50 ,ug of DNA per 10-cm tissue culture dish:40 ,ug of the Kb plasmid and 10 ,g of a coselection marker,the pMClneo-polyA+ plasmid (Stratagene). Transfectantswere selected in 1 mg of G418 per ml (GIBCO/BRL). TheG418-resistant cells were stained with an anti-Kb monoclonalantibody (see below) and Kb-expressing COS-7 cells wereisolated by fluorescence-activated cell sorting (FACS).

Antibodies, Antigens, and Plasmids. The monoclonal anti-bodies 500A2 (anti-CD3e) (11), MR9-4 (anti-V135) (12), 3.155(anti-CD8) (13), 28-13-3S (anti-Kb) (14), and HK 2.1 (anti-Thyl) (15) were kindly provided by J. P. Allison (Universityof California, Berkeley). The second-step antibody used withall monoclonal antibodies, with the exception ofHK 2. 1, waspolyclonal goat anti-mouse fluorescein isothiocyanate (Or-ganon Teknika-Cappel). Binding of HK 2.1 was detectedusing polyclonal goat anti-mouse phycoerythrin (Biomedia,Foster City, CA), because lacZ-mediated conversion offluorescein di-,3-D-galactopyranoside (FDG) to fluoresceinwas being measured in the same experiment.The synthetic peptide analog of hen egg lysozyme (HEL)

amino acids 74-88 (sequence: NH2-Asn-Leu-Ala-Asn-Ile-Pro-Ala-Ser-Ala-Leu-Leu-Ser-Ser-Asp-Ile-OH) has been de-scribed (16). Cyanogen bromide-digested OVA (CNBr OVA;grade IV Sigma) was prepared according to Moore et al. (17).The COS expression vector (pCDM8) (18) and plasmids

containing the full-length OVA cDNA (pAc-Neo-OVA) (17)and the human CD3 ; chain (pCD30 (19) were kindlyprovided by B. Seed (Massachusetts General Hospital, Bos-ton), M. Moore (University of Southern California MedicalSchool, Los Angeles), and L. Lanier (Becton Dickinson),respectively. The pCDM8-OVA plasmid was constructed byfirst subcloning the OVA cDNA, a 1.8-kilobase (kb) EcoRIfragment, into the BlueScript(+) vector (Stratagene). Usingflanking sites present in the BlueScript polylinker, the OVAcDNA was excised as a HindIII/Not I fragment and religatedin the correct transcriptional orientation into the HindIII/NotI cloning sites of the pCDM8 vector.

T-Cell Activation Assays. lacZ/FDG assay. Individual cul-tures containing 5 x 105 APCs, 1 x 106 T cells, and theindicated concentration of antigen were set up in 24-welltissue culture plates (total volume = 1 ml per well). Cultureswere incubated for 6 hr and then harvested for analysis byvigorous pipetting. To distinguish T cells from contaminatingAPCs, the samples were stained for Thyl with the monoclo-nal antibody HK2.1 (see above). The cells were then loadedwith FDG (Molecular Probes) and analyzed for lacZ expres-sion on a flow cytometer as described (2, 5). Analysis wasconfined to the phycoerythrin-stained (Thyl positive) popu-lation, to exclude fibroblasts from the analysis. The meangreen fluorescence of unstimulated, lacZ-negative T cells

was set arbitrarily in channel 85 of a 256-channel histogram.Each histogram shown represents an analysis of 10,000 cells.

lacZ/S-bromo4-chloro-3-indolyl 3-D-galactopyranoside(X-Gal) assay. Individual cultures containing 3 x 104 APCs,3 x 104 T cells, and peptide antigen (if indicated), in a totalvolume of 0.2 ml, were set up in 96-well tissue culture plates.Cultures were incubated for 6 hr and then fixed and stainedfor lacZ expression with X-Gal (6). The plates were washedonce with 0.1 ml of 10 mM phosphate-buffered saline (PBS)per well and then fixed for 5 min on ice with 0.1 ml of ice-cold2% formaldehyde/0.2% gluteraldehyde in PBS per well.After another wash with 0.1 ml ofPBS per well, the cells wereoverlaid with a solution of 1 mg of X-Gal per ml (Sigma), 5mM potassium ferrocyanide, 5 mM potassium ferricyanide,and 2 mM MgCl2. The plates were examined microscopicallyfor the presence of blue (lacZ expressing) cells after anovernight incubation at 370C.

Transient Expression Assays. The DEAE-dextran transfec-tion protocol described by Seed and Aruffo (18) was adaptedfor 96-well plates. KbCOS cells were plated out in 96-welltissue culture plates at a density of 3 x 104 cells per well inRPMI 1640 medium containing 10%o NuSerum (CollaborativeResearch), 100 gg of DEAE-dextran per ml (Pharmacia), 100gM chloroquine diphosphate (Sigma), and the indicatedquantity of DNA. After a 2-hr incubation at 370C, thetransfection mixture was removed, the cells were washedonce with 0.1 ml of PBS per well, and then treated for 2 minwith a solution of 7.5% dimethyl sulfoxide (DMSO) in PBS.The DMSO/PBS solution was then replaced with 0.1 ml ofRPMI 1640 medium/10% fetal bovine serum. The cells wereincubated for 48 hr to allow expression of the introducedgenes. After 48 hr, T cells were added and T-cell activationassays were performed as described above.

RESULTSGeneration of an IL-2-Secreting, lacZ-Inducible T-Cell Hy-

bridoma Specific for the OVA/Kb Complex. Previous exper-iments with JK12/90.1, a class II-restricted T-cell hybrid-oma, had demonstrated that the NFAT-lacZ reporter genecould be used to detect activation of individual T cells inmixed cultures ofT cells and APCs (2). Not only did the lacZassay allow detection of small percentages of activated Tcells but it also permitted the concominant detection of rareAPCs. Additional experiments with JK12/90.1 demonstratedthat ligand-expressing APCs present at frequencies as low as2% could easily be detected by flow cytometry by assayingfor lacZ-expression in cocultured T cells (data not shown).

Cytotoxic T cells, however, do not normally secrete IL-2and were therefore not expected to induce the lacZ reporterconstruct, which is driven by the NF-AT element of the IL-2gene. To overcome this limitation, we used a strategy devel-oped earlier for generating IL-2-secreting, class I-restrictedT-cell hybridomas (20, 21). As detailed in Material andMethods, we generated somatic hybrids by fusing B3 (7), anOVA/Kb-specific cytolytic T-cell clone, with Z.8, a deriva-tive of BW5147 stably transfected with the NF-AT reporterconstruct. CD8 expression, which was absent in the T-cellhybrids, presumably due to fusion with the BW5147-derivedfusion partner, was restored by infecting with a retroviralvector carrying the CD8a gene (8). The level ofTCR and CD8expression on the surface of the transfected cells was deter-mined by staining with monoclonal antibodies. As shown inthe FACS profiles in Fig. 1A, the level of CD3, Vp5, and CD8expression by the B3xZ.8 Lyt2+ T-cell hybrids was compa-rable to that of the parent B3 line.

Individual B3XZ.8 T-cell hybrids were tested for theirability to induce lacZ expression by coculturing them withKb-expressing L cells and CNBr OVA. The FACS profiles inFig. 1B show a comparison of lacZ induction in the TCR-

Immunology: Karttunen et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

0

6022 Immunology: Karttunen et al.

A

00

X en

cnma

100

CD3 VSS c

:- \ /

2nd .

I' I

." t . I^

B

CD$

I

I I

I'

CD3 VaSXAJ~~~CD2nd

log fluorescence

Kb APC

/ Kb APC+ CNBr OVA

/Kb AjC Kb APC

+ CNBr OVA

1. I

f %

I.. ...

101 102 103lacZ activity

FIG. 1. The T-cell hybrid B3xZ.8 (B3Z) expresses CD3, V.35,and CD8 and induces lacZ in the presence of APCs expressing theOVA/Kb ligand. (A) B3, Z.8, and B3Z cells were stained withmonoclonal antibodies 500A2 (anti-CD3) (-), MR9-4 (anti-V85)( ), and 3.155 (anti-CD8) (- - -). Background staining was deter-mined by staining the cells with the second-step antibody, goatanti-mouse fluorescein isothiocyanate, alone (... ). (B) lacZ expres-sion was quantified by coculturing 106 Z.8 cells (middle panel) or B3Zcells (bottom panel) with 5 x 105 Kb-expressing L cells (Kb APC) inthe presence of 100 ,ug of CNBr OVA per ml. After 4 hr the cultureswere harvested and lacZ expression was measured by FACS.

negative parent cell, Z.8 (middle panel), and a lacZ-induciblehybrid, B3xZ.8.14 (abbreviated B3Z) (bottom panel). After6 hr of coculture with Kb-expressing L cells and 100 pug ofCNBr OVA per ml, 80% of the B3Z cells were lacZ positive,as assessed by FACS analysis of FDG-stained samples. Asexpected, no lacZ induction was detected among the TCR-negative Z.8 cells. No lacZ induction was detected in theabsence of the CNBr OVA (Fig. 1B) or when B3Z cells werecocultured with L cells, which did not express the Kb_restricting element (data not shown). lacZ induction by B3Zparalleled IL-2 secretion; both responses varied with theconcentration of the antigenic peptide and were inhibited byeither anti-CD8 or anti-Kb antibody (data not shown).These results demonstrate that lacZ expression by B3Z is

specifically induced by the OVA/Kb ligand. They also showthat the NFAT-lacZ reporter construct can be used tomeasure activation of T-cell hybrids specific for MHC classI- and class II-bound peptides.

Microscopic Detection of lacZ+ T Cells by X-Gal StainingIncreases the Sensitivity at Which Ligand-Expressing APCsCan Be Detected. The minimum number of lacZ1 T cells, andtherefore the minimum number of ligand-expressing APCs,that can be detected by FDG staining is limited by thesensitivity of the FACS and the number of cells required forflow cytometry. In our hands, the limit of detection in aroutine analysis (10,000-20,000 cells) is -3-5% lacZ+ cells.To develop a method suited to the analysis of a large numberof samples, lacZ expression was visualized by staining thecocultured cells with X-Gal (6). X-Gal is cleaved by lacZ toyield 5,5'-dibromo-4,4'-dichloroindigo, which stains thelacZ-expressing cells dark blue.

The photomicrographs of X-Gal-stained samples shown inFig. 2 illustrate how activated, lacZ-expressing T cells be-come strikingly visible as darkly staining cells, readily dis-tinguishable against a background of unstained, lacZ-negative cells (nonactivated T cells and APCs). Becausevisualization of lacZ activity by X-Gal staining is not limitedby the total number of cells required for FACS analysis, thisassay is in effect a far more sensitive screen for T-cellactivation-as few as 1:105 activated T cells can be routinelydetected, a 1000-fold improvement over the FACS assay.Because X-Gal staining requires a visual count of blue-

stained cells, it is a much less quantitative assay of lacZexpression than the FDG/FACS assay. However, as a semi-quantitative or qualitative assay, X-Gal staining is an exquis-itely sensitive indicator of lacZ-expressing T cells and there-fore of ligand-expressing APCs. To determine the minimumfrequency of ligand-bearing APCs that can be detected withthe X-Gal assay, we carried out a limiting dilution analysiswith cells stably expressing the OVA/Kb ligand (K89-7.OVA). These cells process and present OVA and readilyinduce lacZ expression in cocultured B3Z cells (data notshown). In a series of limiting dilution experiments, K89-7.OVA cells were diluted into cells of the parent line, K89-7,which is Kb positive, but does not express OVA. B3Z cellswere added to the cultures and after 6 hr the samples werefixed and stained for X-Gal expression. By scoring eachcoculture negative or positive on the basis of whether or notit contained blue-stained T cells, the frequency of negativecultures for each concentration of K89-7.OVA was deter-mined. By plotting the logarithm of the fraction of negative

4P d'- i4.

IC;;5-i0960 |

\i, I)'-K\

1'NNBr(AV A

FIG. 2. X-Gal staining of B3Z cells stimulated by OVA/Kb-expressing COS cells. Kb-COS cells (3 x 104) and B3Z cells (3 x 104)were cocultured for 6 hr with (Bottom Right) or without (MiddleRight) 100 Ag of CNBr OVA per ml. The OVA cDNA was titratedinto an irrelevant cDNA encoding the human TCR ; chain (Left andTop Right). Kb-COS cells (3 x 104 per well) were transfected with atotal of 1 ,ug of cDNA, containing the indicated quantity of OVAcDNA and a balance of ; cDNA. After 48 hr, 3 x 104 B3Z cells wereadded to each well. After 6 hr of coculture, the cells were fixed andstained with X-Gal. (x 13.)

Proc. Natl. Acad. Sci. USA 89 (1992)

...-I 'I ,"-I "'21

101 102 101 100

1i L

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

0

Proc. Natl. Acad. Sci. USA 89 (1992) 6023

cultures against the number of K89-7.OVA cells per well(Fig. 3), the minimum detectable number of OVA-expressingcells was found to be -27 K89-7.OVA cells in a total samplesize of 30,000 cells, giving a minimum detection frequency of1:1000. This means that ligand-expressing cells present atfrequencies of 1:103 can be detected with a 95% confidencelimit if samples are screened in triplicate.The same experiment was performed with E.G7-OVA (17),

an OVA-expressing transfectant of EL-4, and with KbCOScells (see below) expressing the OVA protein (Fig. 3). Theminimum frequency at which ligand-bearing cells can bedetected appears to vary from one cell line to another.Significantly, MHC-expressing COS cells can function asvery efficient APCs for murine T cells; in a limiting dilutionanalysis, OVA/Kb-expressing COS cells can be detected ata minimum frequency of 1 in 10,000 (Fig. 3).

Kb-Transfected COS Cells Present Transiently ExpressedOVA to B3Z. In devising a strategy for screening a cDNAexpression library, we took advantage of the well-developedtransient expression system developed in COS cells. COScells are transformed cell lines generated by transfecting theCV-1 simian fibroblast line with an origin-defective mutant ofthe simian virus 40 (SV40) virus (22). These cell lines stablyexpress the SV40 large tumor antigen and support the repli-cation of circular DNA containing the SV40 origin. Becauseplasmid expression vectors containing the SV40 origin arereplicated to high copy number within the nuclei of trans-fected COS cells, very high levels of transient protein ex-pression can be obtained. For this reason, COS cells havebeen used extensively for expression cloning of membraneglycoproteins (18, 20, 23).To generate a COS cell line that could present the OVA/Kb

ligand, COS-7 cells were stably transfected with a $enomicclone of the murine Kb class I MHC molecule l(9). TheKb-expressing COS-7 cells (Kb-COS) were analyzed by flowcytometry and found to express Kb at a level comparable to

,o

0

L.1

0 S 10 15 20 25 30 35 40

OVA-expressing cells, no. per well

FIG. 3. Limiting dilution analysis of OVA/Kb-expressing APCs.OVA/Kb-expressing cells (, K89-7.0VA; *, E.G-7; , OVA+Kb-COS) were titrated into their respective OVA-negative parent cells.B3Z cells (3 x 104) were added to each well and after 6 hr ofcoculture, thp cells were stained with X-Gal and scored for thepresence ofactivated, blue-stained T cells. OVA-expressing Kb-COScells were obtained by transfecting Kb-COS cells with 1 ,ug ofOVAcDNA. The frequency of OVA+Kb-COS cells was estimated from aparallel transfection in which Kb-COS cells were transfected with thelacZ gene and stained with X-Gal. For each dilution of OVA-positivecells per well, the frequency of negative cultures was determined byscoring each replicate well negative or positive on the basis ofwhether or not it contained blue-stained T cells. The logarithm of thefraction of negative cultures was plotted against the number ofOVA-positive cells per well. Linear regression (solid lines) was usedto compute the minimum detectable frequency (dotted lines) ofOVA-positive APCs.

EL-4 and other Kb-expressing cultured cell lines (data notshown). In the presence of CNBr OVA, the KbCOS cellswere able to stimulate lacZ expression in B3Z cells (Fig. 2).The ability of the KbCOS cells to present endogenous

proteins to B3Z was tested by inducing transient expressionof the OVA protein. Transient OVA expression was obtainedby transfecting the Kb-COS cells using the DEAE-dextranmethod with an OVA cDNA construct (pCDM8-OVA). Asshown in Fig. 2, pCDM8-OVA transfected KbCOS cellsprocessed and presented transiently expressed OVA to B3Zcells. This result demonstrates that peptides generated fromtransiently expressed proteins can be presented to T cells andthat COS-7 cells can function efficiently as APCs for classI-restricted murine T cells.OVA cDNA Can Be Detected at an Abundance of 1:104. To

determine the minimum amount of OVA cDNA required togive a positive signal (cluster of blue B3Z cells), pCDM8-OVA was diluted with aCDM8 plasmid containing the humanCD3 rchain (pCD3) (19), an irrelevant protein that does notstimulate B3Z. Kb-COS cells were transfected with a total 10ng of DNA, consisting of various ratios of pCDM8-OVA topCD3 plasmid. As shown in Fig. 2 Top, the frequency ofblueclusters decreases with decreasing amount ofpCDM8-OVA.In a series of independent experiments, we have observedthat as little as 10 pg to 100 fg of the pCDM8-OVA plasmidcould be detected. This detection frequency (1:104-105) iscomparable to those observed in other expression cloningstrategies, such as antibody epitope screens in Agtll or inCOS cells. These results demonstrate that the lacZ/COS cellassay is sufficiently sensitive to allow the detection ofcDNAmessages encoding antigenic peptides at frequencies ex-pected in the primary screen of an expression library.Based on the data presented here, we envision the follow-

ing strategy for the expression cloning of class I-restrictedT-cell antigens. First, lacZ-inducible T-cell hybrids specificfor the unknown ligand would be obtained by fusing aBW5147-derived lacZ+ fusion partner with T-cell clonesspecific for the unknown antigen. Second, a cDNA librarywould be prepared from cells that stimulate these hybrids andare therefore expected to express the mRNA message for theunknown peptide. Individual pools of the cDNA librarywould be transfected into COS cells stably expressing therelevant MHC molecule. After 2 days, the plates would bescreened by adding the lacZ-inducible T-cell hybrids andassaying for lacZ expression. Those pools that gave rise topositive wells would be redivided into smaller pools and theentire process would be repeated until the antigen-encodingcDNA was isolated.

DISCUSSIONIn this paper we describe two important methods for thestudy of T-cell activation. First, we show that the lacZ T-cellactivation assay is a powerful tool for detecting rare ligand-expressing APCs. Second, we demonstrate that COS cellsefficiently express peptides derived from transiently ex-pressed proteins as peptide/MHC complexes on the cellsurface. Taken together, these observations suggest thatunknown T-cell antigens could be identified by an expressioncloning strategy using lacZ-inducible T cells as a screen forligand-expressing APCs. Because COS cells function asefficient APCs, large numbers of cDNA clones can bescreened at one time by performing simple transient expres-sion assays. We demonstrate here that the frequency withwhich antigen-encoding cDNA can be detected (1: 105) iscomparable to those observed in other expression cloningstrategies, such as antibody epitope screens. In fact, thiscloning strategy should be considered the T-cell analog of theantibody epitope screen developed by Aruffo and Seed forthe cloning of membrane glycoproteins.

Immunology: Karttunen et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

0

6024 Immunology: Karttunen et al.

The development of expression cloning systems for T-cellantigens has been hampered by the complexity of the TCRligand. The generation ofthe peptide/MHC ligand in the APCis a complex and poorly understood process (reviewed in ref.24). The antigenic peptides are generated within the APC byproteolysis of native proteins and are subsequently presentedon the cell surface as peptide/MHC complexes. Direct ap-proaches for identifying individual antigenic peptides or theirprecursor proteins using conventional T-cell assays are dif-ficult. The problem of obtaining micromolar quantities of rareproteins for performing T-cell stimulation assays is furthercompounded by the need to generate physiologically activepeptide fragments. This difficulty can be partially circum-vented by purifying naturally processed peptides directlyfrom APCs (25). Using HPLC technology, several abundantendogenous peptides have been successfully isolated andsequenced (26, 27). How successfully this biochemical ap-proach could be applied to antigenic peptides present in smallamounts still remains to be established.

Expression cloning ofantigenic peptides is an alternative tobiochemical purification strategies and is not dependent uponthe total amount of peptide present, for as few as 100-500peptide/MHC complexes are sufficient for stimulating a Tcell (28, 29). Not only does the expression cloning strategyproposed here obviate the need to purify relatively largequantities of protein or peptide, it is also less labor intensiveand requires only commonly available equipment and re-agents.

Expression cloning has been employed to identify genesencoding T-cell antigens on one previous occasion. Fourtumor-specific ("tum-") antigens were identified by screen-ing candidate APCs with antigen-specific cytotoxic T lym-phocytes (30). This accomplishment was particularly remark-able considering that a genomic cosmid DNA library wasscreened by generating several thousand stable transfectants,which were subsequently assayed by measuring their abilityto induce proliferation of individual cytotoxic T lymphocyteclones. Our strategy differs from the one used to isolate thetum- antigens in several important respects. (i) By using atransient expression assay, candidate APCs can be generatedand screened quickly (days versus weeks). (ii) Because thelacZ assay allows the screening of bulk cultures of APCs inwhich the frequency of ligand-expressing cells can be low(1:103), large numbers of candidate APCs can be screenedwith relatively little effort. (iii) By screening cDNA librariesrather than genomic libraries, the frequency of antigen-encoding messages is expected to be higher. (iv) Our ap-proach does not necessitate recloning of the desired genefrom the genome of the ligand-expressing APCs. By using ourproposed strategy, it should be possible to dramaticallyreduce the amount of time and labor required to screen acDNA library and to increase the likelihood of findingantigen-coding sequences.

It should also be noted that the murine Kb-expressing COScells can also serve as a model APC for the study of class Iantigen presentation mechanisms. The ease with which ex-pression constructs can be introduced into COS cells makesthem an attractive alternative to the widely used vacciniaexpression system (31). Transient gene expression can beobtained in a large fraction (20-50%) of COS cells by directtransfer of expression vectors containing the SV40 replica-tion origin, obviating the tedious requirement of constructingintermediate vaccinia recombinants. Also, exogenous geneexpression in COS cells is less likely to be influenced bypotentially complicating effects of the up to 200 other viralgenes expressed by vaccinia recombinants (32).

Finally, it is of obvious interest to determine if the lacZ/COS cell system could also be adapted for use with classII-restricted T-cell antigens. Whether or not the COS cells

can be induced to process and present endogenously ex-pressed peptides on class II MHC remains to be established.

We acknowledge the technical assistance of Mr. Fred Gonzalezand Ms. Elaine Callas. We are also grateful to Drs. M. Bevan, M.Moore, B. Seed, and I. Stroynowski for their gifts of cells or DNAs.This work was supported by grants to N.S. from the NationalInstitutes of Health (AI-26604), the Pew Biomedical Program, andthe Committee on Research, University ofCalifornia, Berkeley. J.K.is a recipient of a Howard Hughes predoctoral fellowship.

1. Sinha, A. A., Lopez, M. T. & McDevitt, H. 0. (1990) Science248, 1330-1338.

2. Karttunen, J. & Shastri, N. (1991) Proc. Natl. Acad. Sci. USA88, 3972-3976.

3. Fiering, S., Northrop, J. P., Nolan, G. P., Matilla, P., Crab-tree, G. R. & Herzenberg, L. A. (1990) Genes Dev. 4, 1823-1834.

4. Shaw, J. P., Utz, P. J., Durand, D. B., Toole, J. J., Emmel,E. A. & Crabtree, G. R. (1988) Science 241, 202-205.

5. Nolan, G. P., Fiering, S., Nicolas, J.-F. & Herzenberg, L. A.(1988) Proc. Nati. Acad. Sci. USA 85, 2603-2607.

6. Sanes, J. R., Rubenstein, J. L. R. & Nicolas, J.-F. (1986)EMBO J. 5, 3133-3142.

7. Hosken, N. A. & Bevan, M. J. (1990) Science 248, 367-370.8. Glaichenhaus, N., Shastri, N., Littman, D. R. & Turner, J. M.

(1991) Cell 64, 511-520.9. Weiss, E., Golden, L., Zakut, R., Mellor, A., Fahrner, K.,

Kvist, R. & Flavell, R. A. (1983) EMBO J. 2, 453-462.10. Ausubel, F. M., Brent, R., Kingston, R. E., Moore, D. D.,

Seidman, J. G., Smith, J. A. & Struhl, K., eds. (1990) CurrentProtocols in Molecular Biology (Greene and Wiley Intersci-ence, New York), Vol. 1.

11. Havran, W. L., Poenie, M., Kimura, J., Tsien, R., Weiss, A.& Allison, J. P. (1987) Nature (London) 330, 170-173.

12. Kanagawa, O., Utsunomiya, Y., Bill, J., Palmer, E., Moore,M. W. & Carbone, F. R. (1991) J. Immunol. 147, 1307-1314.

13. Sarmiento, M., Glasebrook, A. L. & Fitch, F. W. (1980) J.Immunol. 125, 2665-2672.

14. Ozato, K. & Sachs, D. H. (1981) J. Immunol. 126, 317-321.15. Lancki, D. W., Ma, D. I., Havran, W. L.& Fitch, F. W. (1984)

Immunol. Rev. 81, 65-94.16. Shastri, N., Gammon, G., Miller, A. & Sercarz, E. E. (1986) J.

Exp. Med. 164, 882-8%.17. Moore, M. W., Carbone, F. R. & Bevan, M. J. (1988) Cell 54,

777-785.18. Seed, B. & Aruffo, A. (1987) Proc. Natl. Acad. Sci. USA 84,

3365-3369.19. Lanier, L. L., Yu, G. & Phillips, J. H. (1989) Nature (London)

342, 803-805.20. Burgert, H. G., White, J., Weltzien, H. U., Marrack, P. &

Kappler, J. W. (1989) J. Exp. Med. 170, 1887-1904.21. Rock, K. L., Rothstein, L. & Gamble, S. (1990) J. Immunol.

145, 804-811.22. Gluzman, Y. (1981) Cell 23, 175-182.23. Aruffo, A. & Seed, B. (1987) Proc. Natl. Acad. Sci. USA 84,

8573-8577.24. Brodsky, F. M. & Guagliardi, L. (1991) Annu. Rev. Immunol.

9, 707-744.25. Rotzschke, O., Falk, K., Deres, K., Schild, H., Norda, M.,

Metzger, J., Jung, G. & Rammensee, H. G. (1990) Nature(London) 348, 252-254.

26. Falk, K., Rotzschke, O., Stevanovic, S., Jung, G. & Ramm-ensee, H. G. (1991) Nature (London) 351, 290-2%.

27. Jardetzky, T. S., Lane, W. S., Robinson, R. A., Maden, D. R.& Wiley, D. C. (1991) Nature (London) 353, 326-329.

28. Christinck, E. R., Luscher, M. A., Barber, B. H. & Williams,D. B. (1991) Nature (London) 352, 67-70.

29. Falk, K., Rotzschke, O., Deres, K., Metzger, J., Jung, G. &Rammensee, H. G. (1991) J. Exp. Med. 174, 425-434.

30. Boon, T., Van Pel, A., De Plaen, E., Chomez, P., Lurquin, C.,Szikora, J., Sibille, C., Mariame, B., Van den Eynde, B. &Lethe, B. (1989) Cold Spring Harbor Symp. Quant. Biol. 54,587-5%.

31. Chakrabarti, S., Brechling, K. & Moss, B. (1985) Mol. Cell.Biol. 5, 3403-3420.

32. Moss, B. (1990) Annu. Rev. Biochem. 59, 661-688.

Proc. Natl. Acad. Sci. USA 89 (1992)

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

0