of December 20, 2015.This information is current as

I MHCNK Cells Can Recognize Different Forms of Class

and Richard G. MillerRuey-Chyi Su, Sam K. P. Kung, Jean Gariépy, Brian H. Barber

http://www.jimmunol.org/content/161/2/7551998; 161:755-766; ;J Immunol 

Referenceshttp://www.jimmunol.org/content/161/2/755.full#ref-list-1

, 30 of which you can access for free at: cites 53 articlesThis article

Subscriptionshttp://jimmunol.org/subscriptions

is online at: The Journal of ImmunologyInformation about subscribing to

Permissionshttp://www.aai.org/ji/copyright.htmlSubmit copyright permission requests at:

Email Alertshttp://jimmunol.org/cgi/alerts/etocReceive free email-alerts when new articles cite this article. Sign up at:

Print ISSN: 0022-1767 Online ISSN: 1550-6606. Immunologists All rights reserved.Copyright © 1998 by The American Association of9650 Rockville Pike, Bethesda, MD 20814-3994.The American Association of Immunologists, Inc.,

is published twice each month byThe Journal of Immunology

by guest on Decem

ber 20, 2015http://w

ww

.jimm

unol.org/D

ownloaded from

by guest on D

ecember 20, 2015

http://ww

w.jim

munol.org/

Dow

nloaded from

NK Cells Can Recognize Different Forms of Class I MHC1

Ruey-Chyi Su,* Sam K. P. Kung,* Jean Gariepy,* Brian H. Barber,† and Richard G. Miller* 2

NK recognition and lysis of targets are mediated by activation receptor(s) whose effects may be over-ridden by inhibitory receptorsrecognizing class I MHC on the target. Incubation of normal lymphoblasts with a peptide that can bind to their class I MHCrenders them sensitive to lysis by syngeneic NK cells. By binding to class I MHC, the peptide alters or masks the target structurerecognized by an inhibitory NK receptor(s). This target structure is most likely an “empty” dimer of class I heavy chain andb2mas opposed to a “full” class I trimer formed by binding of specific peptide that is recognized by CTL. The Journal of Immunology,1998, 161: 755–766.

I t is now widely accepted that NK cells recognize and lysetarget cells through the interplay of two families of receptors(1–3). Activating receptors, when occupied, trigger lysis of

the target cell being recognized. The activating signal, however,can be overridden by a dominant negative signal from an inhibi-tory receptor when the latter interacts with its ligand (if present) onthe target cell. The ligand(s) for activating receptors remains un-known, but there is a general agreement that the ligand for some(perhaps all) inhibitory receptors is associated with MHC class Ialleles, with a particular receptor being specific for a limited num-ber of class I alleles. C1R is an HLA-A, HLA-B null human tumorcell line sensitive to lysis by polyclonal human NK cells. Storkuset al. found that transfection of some (but not all) HLA-A orHLA-B molecules into C1R protected it from NK lysis (4). Byperforming exon shuffling and point mutation experiments,Storkus et al. showed that thea1-a2 region of MHC class I ap-pears to be critical in determining the specificity of MHC class I asan inhibitory ligand (5), and that the amino acids in the peptidebinding site of MHC class I molecules appear to be important inthe protection (6). In addressing whether occupation of the peptidebinding site was important, Storkus et al., using C1R cells trans-fected with protective human HLA-A or HLA-B MHC class Imolecules, found that addition of peptide that could bind to a pro-tective MHC class I reversed protection, i.e., sensitivity to lysiswas restored upon peptide binding (7). Similar observations havebeen obtained in a more physiologic setting in which normal, un-transformed lymphoblasts and syngeneic (polyclonal) mouse NKcells were used, respectively, as target and effector cells (8, 9).They found that the lymphoblasts, which are resistant to lysis bysyngeneic mouse NK cells, could be rendered sensitive to lysis ifpeptides that could bind to the MHC class I of the normal cellswere included in the assay. Eight peptides, capable of binding Kb,Db, Kd, or Ld class I molecules, were tested. All eight peptides

tested (seven of which included CTL epitopes and one of whichdid not) could sensitize normal targets for lysis if they could bindto the class I of the target, but otherwise had no effect (9). Onepossible explanation of these results, consistent with those ofStorkus et al. (7), is that binding of peptide to MHC class I isaltering or masking an inhibitory ligand recognized by an inhibi-tory receptor and thus sensitizing the cells to lysis. Identification ofhuman KIRs3 (p58.1, p58.2, or p70) and murine Ly49A moleculesas NK inhibitory receptors specific for particular MHC class Ialleles facilitates a detailed study of specific receptor-ligand inter-actions. Ly49A is known to recognize Dd (10, 11), and recentevidence indicates that recognition requires that Dd is loaded withpeptide (12, 13). Both groups used mutant cell lines lacking func-tional peptide transporter molecules so that only empty (and un-stable) MHC class I molecules appear on the cell surface. Thesecan bind and be stabilized by high affinity class I-binding exoge-nous peptide (14, 15). Both groups used Ly-49A1 NK cells aseffector cells and the mouse mutant cell lines RMA-S (12) andLKD8 (13), transfected with Dd as target cells. Addition of peptidethat could bind to Dd was shown to protect the cell lines from lysisby Ly49A1 mouse NK cells. The extent of protection correlatedwith the extent to which the added peptide stabilized Dd expression(12). Both groups suggested that the role of peptide was to promotethe assembly and cell surface expression of MHC class I and thatthere was no peptide specificity in Ly49A recognition of the Dd

molecule. In a similar study, Malnati et al. used RMA-S cellstransfected with HLA-B27 as targets, human NK clones express-ing KIR receptors specific for HLA-B27 as effectors, and exoge-nous synthetic peptide ligands of HLA-B27 to stabilize surfaceexpression of the HLA molecules on RMA-S cells (16). One of thefour peptide ligands specific for HLA-B27 tested provided protec-tion from lysis by the specific NK clones (16). The protection wasindependent of the peptide binding affinity to HLA-B27. By per-forming further analysis of HLA-B27-specific peptides usingamino acid substitutions, Peruzzi et al. found that the side chainsof the seventh and eighth amino acids of “protective” peptides

*Department of Medical Biophysics, Ontario Cancer Institute, and†Department ofImmunology, University of Toronto, Toronto, Ontario, Canada

Received for publication December 19, 1997. Accepted for publication March24, 1998.

The costs of publication of this article were defrayed in part by the payment of pagecharges. This article must therefore be hereby markedadvertisementin accordancewith 18 U.S.C. Section 1734 solely to indicate this fact.1 This work was supported by a Connaught-University of Toronto Research Fundgrant and a Human Frontiers of Science grant (both to R.G.M.). S.K.-P. Kung is aresearch student of the National Cancer Institute of Canada supported by funds pro-vided by the Canadian Cancer Society.2 Address correspondence and reprint requests to Dr. Richard G. Miller, Departmentof Medical Biophysics, Ontario Cancer Institute, 610 University Ave., Toronto, On-tario, Canada M5G 2M9.

3 Abbreviations used in this paper: KIR, human killer cell inhibitory receptor (e.g.,p58.1, p58.2, or p70); H, heavy chain of MHC class I; pH, MHC type I-specificpeptide with high binding affinity; Flu-NP-Db, restricted epitope of influenza nucleo-protein, ASNENMETM; Flu-NP-Kd, restricted epitope of influenza nucleoprotein,TYQRTRALV; Tum2, Ld-restricted epitope, ISTQNHRALDLVAAK; HIVp-Dd, re-stricted epitope of human immunodeficiency virus gp160, RGPGRAFVTI; BFA,brefeldin A; OVApX-bio, biotinylated OVAp, bio-XSIINFEKL; OVApK-bio, biotinyl-ated OVAp, SIINFEK(bio)L; GAD-Flu-NP, Flu-NP with three additional amino acidsadded to the N-terminus, GADASNENMETM; OVAp-Kb, restricted epitope ofchicken ovalbumin, SIINFEKL; pL, MHC class I-specific peptide with low bindingaffinity.

Copyright © 1998 by The American Association of Immunologists 0022-1767/98/$02.00

by guest on Decem

ber 20, 2015http://w

ww

.jimm

unol.org/D

ownloaded from

were conserved and may be involved in NK recognition (17). Thisinvolvement may be either indirect, by affecting the conformationof the KIR binding site, or direct, through interference with KIRbinding to the class I heavy chain (18).

In summary, binding of peptide to MHC class I has been shownto sensitize targets to NK lysis (7–9), as well as to protect targetsfrom NK lysis (12, 13, 16, 17). We here try to reconcile theseapparently contradictory findings by assessing the possibility thatNK cells can recognize different forms of MHC class I molecules.There are four possible forms of MHC class I molecules expressedon the normal cell surface. The majority exist as trimolecular com-plexes, each composed of a properly folded heavy chain (H) con-taining the peptide binding groove, a noncovalently associatedb2m molecule, and a peptide (p) that can bind to MHC class I withhigh affinity (therefore, pH) in the peptide-binding groove (thuspH-H-b2m) (19). Three other unstable forms of MHC class I,b2m-H, pH-H, and H (perhaps in decreasing order of stability) canbe found (19, 20). In addition, pL-H-b2m molecules in which thepeptide is either too long or lacks the proper binding motif and thusbinds with low affinity (therefore, pL) are probably also present.For the cell line RMA-S (H-2b), H-b2m and H have been showndirectly to have half-lives of,30 min, and at least one particularpL-H-b2m has been inferred indirectly to have a comparably shortlifetime and is likely to give rise to H-b2m, whereas pH-H-b2mappears to have a lifetime much greater than 4 h (19). Only two ofthese four forms of MHC class I molecule (the trimolecular com-plex of H chain,b2m and peptide bound with high affinity, and thebimolecular complex in which the peptide is not present) are likelyto be expressed in appreciable numbers on the surface of a normalcell (14, 15, 19–21). Approximately 10% of Db molecules ex-pressed on the cell surface are likely to be bimolecular MHC classI (H-b2m) molecules because 1) about 10% of Db molecules onEL4 cells can be bound very rapidly by exogenous peptide (half-time of 9.36 1.1 min at 37°C) (14); and 2) the binding of exog-enous peptide to purified Db H chain was determined to have ahalf-time of 13 h, presumably because most of it was denatured,while binding of peptide to purified H-b2m bimolecules had ahalf-time of ,10 min at 22°C (20). Although both H chain andH-b2m bimolecule can potentially bind exogenous peptide, addedpeptide is most likely to bind to H-b2m bimolecular MHC class Ibecause H chain is very unstable at 37°C. We therefore refer toH-b2m bimolecular MHC class I as empty MHC class I. All fourforms of MHC class I molecule, but particularly pH-H-b2m andempty H-b2m molecules (because of their appreciable abundanceon the cell surface), might be recognized by NK inhibitory recep-tors involved in self-recognition.

In this report, we first reinvestigate and further characterize theexperimental system developed by Chadwick et al. in which incu-bation of normal lymphoblasts with class I binding peptide sensi-tizes them to lysis by syngeneic NK cells (8, 9). We conclude thatthe peptide is most likely altering or masking the ligand recognizedby an inhibitory receptor. This ligand appears to be empty MHCclass I, as defined above. Second, to reconcile this conclusion withthe fact that the inhibitory receptor Ly49A recognizes the trimo-lecular complex of class I Dd plus peptide, we have investigatedthe lysis of normal lymphoblasts by syngeneic Ly49A1 andLy49A2 NK cells in the presence and the absence of class I bind-ing peptide. The results are consistent with the conclusion thatLy49A recognizes Dd plus peptide but, at the same time, suggestthat there are additional inhibitory receptors that recognize emptyMHC class I molecules. We propose a model in which a smallchange in the total inhibitory signal delivered by several inhibitoryreceptors can switch a cell from resistance to sensitivity to lysis byNK cells.

Materials and MethodsNK generation

The method used for producing activated NK cells (LAK cells) was iden-tical with that used previously (8, 9, 22). Briefly, 23 106 nylon woolnonadherent spleen cells from B6 athymic nude mice (The Jackson Lab-oratory, Bar Harbor, ME) were cultured at 37°C for 3 to 4 days in 5 ml ofa-MEM supplemented with 10% FCS, 50mM 2-ME, and 10 mM HEPESbuffer (hereafter referred to as CM), containing 500 U/ml mouse rIL-2. Insome experiments, as specified, B6 CD8 knockout mice (23), BALB/cathymic nude mice (The Jackson Laboratory), or normal BALB/c mice(The Jackson Laboratory) depleted of T cells using anti-CD4/CD8 Abs andDynabeads (Dynal, Oslo, Norway), all depleted of nylon wool adherentcells, were used. Mouse rIL-2 was obtained as a supernatant from a cellline transfected with the IL-2 gene (24). These cells were cultured in 25-cm2 flasks at 37°C in a 10% CO2 in air incubator. Yields typically ex-ceeded 5000 U/ml of rIL-2.

Target cell generation

Target cells were B6 Con A (ICN Pharmaceuticals Canada, Montreal, Can-ada) blast cells produced by incubating 107 B6 splenocytes for 3 days in 10ml of CM supplemented with Con A (2mg/ml). On day 3, Con A blast cellswere harvested on Lympholyte M (Cedarlane, Hornby, Canada) and51Crlabeled by incubating about 63 106 cells for 90 min at 37°C with 360mCiof Na51CrO4 (New England Nuclear, Boston, MA) in 150ml of PBS con-taining 67% FCS. They were then washed three times with CM containing1% FCS to remove nonincorporated Na51CrO4.

MHC class I binding peptides

The effect of MHC class I binding peptides on normal lymphoblastsensitivity to NK lysis was assessed by pulsing lymphoblasts with theexperimental peptide (at a concentration of 1 ng/ml in CM unless statedotherwise) for 45 min at 4°C before the assay. The peptides used werea Db-restricted epitope of influenza nucleoprotein, ASNENMETM,(Flu-NP366 –374) (25), a Kb-restricted epitope of chicken OVA,SIINFEKL, (OVAp258 –265) (26); a Dd-restricted epitope of HIV gp160,RGPGRAFVTI, (HIVp318–327) (27); a Kd-restricted epitope of influenzanucleoprotein, TYQRTRALV, (Flu-NP147–155) (25, 28); and an Ld-re-stricted epitope referred to as Tum2, ISTQNHRALDLVAAK,(Tum-12–26) (29). Both Flu-NP peptides (.90% purity) were synthesizedand purified by the Alberta Peptide Institute (Edmonton, Canada). ChickenOVA, SIINFEKL (OVAp258–265), and its derivatives, biotinylated OVApeptide, (bio)-XSIINFEKL, where X is aminocaproic acid (a linker be-tween biotin and the peptide), and SIINFEK(bio)L were prepared by theOntario Cancer Institute Biotechnology Laboratory, using an Applied Bio-systems Peptide Synthesizer (Applied Biosystems, Foster City, CA). HIVp(.90% purity) was a gift from Dr. D. Williams (University of Toronto).Flu-NP(Db), OVAp, HIVp, and Flu-NP(Kd) peptides are natural ligands forDb, Kb, Dd, and Kd, respectively, and bind to Db, Kb, Dd and Kd with highaffinities (25–28). Tum2 peptide binds specifically to Ld molecules afterbeing processed to its optimal length by proteases in serum. During thepulsing condition used in the current study, unprocessed Tum2 peptidecannot bind to Ld (9).

Peptide pulsing and cytotoxicity assay

Methods for measuring lytic activity were identical with those used pre-viously (9, 22). After three washes,51Cr-labeled Con A lymphoblasts wereincubated with peptide in 3 ml of CM for 45 min at 4°C and washed againbefore being used in a 4.5-h51Cr release assay performed in 96-well V-bottom microtiter plates using 2000 targets/well, dispensed in 100-ml ali-quots and effectors at an E:T cell ratio as indicated or at 30:1, also addedin 100-ml aliquots. For experiments where preincubation of NK cells withF(ab9)2 anti-Ly49A mAb (JR9-318) was required, the preincubation wasperformed at 37°C for 30 to 45 min while preparing target cells for theassay. Specific lysis was calculated as % specific lysis5 (E 2 S)/(T 2 S)3 100, where each value represents the mean6 SE of five replicates. E isthe experimental mean of51Cr released, S is the amount of51Cr releasedwhen the target cells were cultured in medium alone, and T is the totalamount of51Cr released in the presence of 2% acetic acid. Dialyzed FCS(12 kDa cutoff) was regularly used in place of regular FCS during the51Crlabeling, pulsing, and assay stages (14, 30).

CTL generation and maintenance

Generation of peptide-specific CTL was performed as described previously(31). Briefly, lymphocytes from normal C57BL/6 (B6) mice were depletedof B cells by passage through nylon wool and cultured at 5 to 63 106

756 CLASS I BINDING PEPTIDES AND NK RECOGNITION

by guest on Decem

ber 20, 2015http://w

ww

.jimm

unol.org/D

ownloaded from

cells/ml in 10 ml of CM in the presence of 1 ng/ml of peptide (Flu-NP orOVAp) and 5 U/ml of mouse rIL-2 (24). On day 7, CTL were harvested onLympholyte M (Cedarlane) and used in the cytotoxicity assay. To maintaina CTL line, 106 cells were harvested after 7 to 10 days of culture andcultured with 23 106 irradiated (15 gray) B6 spleen cells in the presenceof 1 ng/ml of peptide and 5 U/ml of mouse rIL-2 as described above.

Cold target competition assay

B6 radiolabeled Con A lymphoblasts, either pulsed or unpulsed with pep-tide (1 ng/ml), were tested as targets using either B6 NK cells or peptide-specific B6 CTL lines as effectors, as described in the cytotoxicity assayexcept that unlabeled B6 Con A lymphoblasts, either pulsed (1 ng/ml) orunpulsed, were included in the wells at zero-, one-, three-, or fivefoldmultiplicities of the labeled targets as indicated. Cold and hot targets werepremixed before the addition of effector cells (i.e., NK cells or CTL lines).A 4.5-h 51Cr release assay was performed in 96-well V-bottom microtiterplates, and specific51Cr release was measured. Specific lysis was calcu-lated as described in the cytotoxicity assay section.

Conjugate formation assay

FITC (green dye, Sigma, St. Louis, MO)-labeled LAK cells were preparedas described by Kung et al. (32). Briefly, day 3 or day 4 B6 LAK cells(10–123 106) were incubated with a FITC solution (10mg/ml PBS, finalconcentration) at 37°C for 18 min. Excess FITC was removed by centri-fuging the cells through 5 ml of 6% BSA/PBS. The cells were then washedtwice with 1% BSA/PBS. PKH26 (red dye, Sigma)-labeled target cellswere prepared according to the manufacturer’s protocol. Briefly, YAC-1and B6 Con A lymphoblasts were washed twice with serum-free mediumand then incubated with PKH26 dye (43 1026 M) in labeling buffer(diluent C; 107 cells/ml) at 25°C for 3 to 5 min. The staining reaction wasstopped by adding an equal volume of 1% BSA/PBS. The cells werewashed three times with 10% CM to remove excess PKH26 dye. Theconjugation formation assay used was described by Cavarec et al. (33).FITC-labeled LAK cells were pelleted and incubated with the PKH26-labeled target cells (B6 Con A lymphoblasts, B6 Con A lymphoblastspulsed with OVAp peptide, or YAC-1) at an E:T cell ratio of 3:1 for 10 minat 37°C. At the end of incubation, the effector-target mixture was resus-pended in 1 ml of 1% BSA/PBS and kept at 4°C before being analyzed forits fluorescence. For negative controls, LAK cells and target cells, at an E:Tcell ratio of 3:1, were mixed and vortexed without any cocentrifugationbefore the analysis with FACScan.

Time delay experiment with or without brefeldin A (BFA)

B6 Con A lymphoblasts were pulsed with peptide and washed free ofunbound peptide as described in the cytotoxicity assay section. The cellswere then incubated in CM at 37°C with or without BFA (5mg/ml; Sigma-Aldrich Canada, Oakville, Canada) for varying lengths of time before be-ing tested as targets in a 4.5-h51Cr release assay using either syngeneic B6NK cells or peptide specific B6 CTL lines.

Flow cytometry/FACS analysis

To measure newly emerged empty MHC class I molecules, day 3 ConA-activated lymphoblasts were prepulsed with nonlabeled OVAp peptide(10 ng/ml) for 45 min to fill empty Kb molecules, washed free of unboundpeptide, and then incubated at 37°C for 0 or 90 min in the presence or theabsence of BFA before being pulsed with biotinylated OVAp peptide (100ng/ml). To measure the effect of BFA on the existing empty MHC class Imolecules, day 3 Con A-activated lymphoblasts were incubated at 37°C for0 or 4 h in thepresence or the absence of BFA before being pulsed withbiotinylated OVAp peptide (100 ng/ml). FITC-conjugated mAb 5F1, pu-rified from the hybridoma 5F1-2-14 (34), was used to detect the expressionof peptide-Kb complexes on the cell surface immediately after the pulsingwith biotinylated OVAp peptide, as it has been shown (34) that this mAbdoes not recognize empty Kb molecules. OVApX-bio (does not bind toMHC class I) and OVApK-bio were used in the staining assay. The bindingof biotinylated OVAp was visualized with R-phycoerythrin-conjugatedstreptavidin (Sigma), which binds to biotin, and analyzed using the LYSISII program (Becton Dickinson, Mountain View, CA).

Cell sorting for Ly49A1 and Ly49A2 NK subsets

Day 3 BALB/c LAK cultures were harvested and resuspended in 1% BSA/PBS (107 cells/ml). The cells were then incubated with 4mg of JR9-318mAb (35) (obtained from Dr. D. Raulet with the permission of Dr. J.Roland, Pasteur Institute, Paris, France) per 106 cells at 4°C on a rotator for45 min. JR9-318 mAb recognizes the NK inhibitory receptor, Ly49A (35).Stained cells were washed with cold 1% BSA/PBS and then incubated with

sheep anti-mouse IgG conjugated to Dynabeads (Dynal; one bead per cell)for 45 min at 4°C on a rotator. Ly49A1 cells, bound to the magnetic beads,were separated from Ly49A2 cells, and both were cultured in 5 ml of CMcontaining 500 U/ml mouse rIL-2 (24) for an additional 3 to 4 days asdescribed above. Ly49A1 cells that were bound to the beads dissociatedfrom the beads during the overnight incubation, and the beads were thenremoved.

F(ab9)2 fragment generation

For F(ab9)2 fragment generation, 2 mg of affinity-purified anti-Ly49A mAb(JR9-318) was resuspended in 1 ml of 0.1 M sodium citrate buffer (pH 3.5)and then digested with 10mg of pepsin (Boehringer Mannheim, Mann-heim, Germany) at 37°C for 4 to 5 h. The reaction was stopped by adding0.5 vol of 1 M Tris to the mixture. After a centrifugation at 10,000 rpm for30 min, the supernatant was collected and mixed with protein A-Sepharosebeads (Sigma, St. Louis, MO) to remove undigested Ab and Fc fragments.The purity and the binding activity of F(ab9)2 fragments were checked by10% SDS-PAGE and flow cytometry, respectively.

ResultsNormal Con A lymphoblasts become sensitive to lysis mediatedby syngeneic NK cells after being pulsed with MHC class Ibinding peptide

The effect of pulsing normal Con A lymphoblasts with MHC classI binding peptide was studied. Effector cells were splenocytes fromB6 (H-2 Kb, Db) athymic nude (T cell-deficient) or CD8 knockoutmice (23) depleted of B cells by passage through nylon wool andcultured for 3 to 4 days in a high concentration of mouse rIL-2 (8,22). This procedure produces a population of highly enriched andactivated NK cells (often referred to as LAK or lymphokine-acti-vated killer cells). Target cells were B6 Con A lymphoblastspulsed for 45 min at 4°C with the Db-binding peptide Flu-NP366–374

(Flu-NP) (25), the Kb-binding peptide OVA258–265(OVAp) (26), orthe Db-binding peptide GAD-Flu-NP (Flu-NP with three additionalamino acids added to the N-terminus of the Flu-NP366–374peptide).Both Flu-NP and OVAp are natural ligands of an optimum lengththat can bind with high affinity to Db or Kb MHC class I, respec-tively, in ,30 min (36). In contrast, GAD-Flu-NP peptide mightbind to Db MHC class I with a relatively low affinity (14, 36). Withvarying E:T cell ratios, significant lysis of Flu-NP- or OVAp-pulsed target cells was always observed for E:T cell ratios of 3:1to 10:1, and usually reached a maximum value at ratios of 10:1 to30:1 (Fig. 1A). Normal Con A lymphoblasts pulsed with mediumalone were, as expected, resistant to NK-mediated lysis. Whennormal Con A lymphoblasts pulsed with varying concentrations ofFlu-NP or OVAp peptide were used in the assay, significant lysisover background of Flu-NP- or OVAp-pulsed target cells was seenfor peptide concentrations as low as 1 pg/ml with the lysis valuesplateauing in the 10 to 100 pg/ml range (Fig. 1B). When normalCon A lymphoblasts were pulsed with a too-long peptide, GAD-Flu-NP, no increase in lysis was observed over the whole dose-response range (Fig. 1B). Pulsing normal Con A lymphoblastswith peptides that could not bind to either Db or Kb MHC class Idid not sensitize these target cells to lysis mediated by syngeneicNK cells (data not shown). Furthermore, no significant lysis ofnormal Con A lymphoblasts was observed when NK cells werepulsed with Flu-NP for 45 min at 4°C and then used as effectorcells (data not shown). Thus, sensitization to lysis required that thetarget cells be exposed to the added peptide and that the addedpeptide have both the correct length and the correct motif to bindto a MHC class I molecule expressed.

The added peptide might sensitize normal lymphoblasts to NKlysis by altering the level of overall MHC class I expression ratherthan through direct binding to MHC class I. It is well establishedthat there is an inverse relationship between sensitivity to NK lysisand MHC class I expression (4, 37). Thus, as little as a twofold

757The Journal of Immunology

by guest on Decem

ber 20, 2015http://w

ww

.jimm

unol.org/D

ownloaded from

decrease in MHC class I can double the amount of lysis observedfor a particular target cell (4). It is possible that binding of theadded peptide to MHC class I on the target cell surface induces NKsensitivity by inducing a relatively modest down-regulation ofMHC class I expression. However, we found, if anything, a slightincrease (,10%) in the level of expression of Kb or Db after B6Con A lymphoblasts were pulsed with OVAp peptide or Flu-NPpeptide, respectively (data not shown). Hence, we conclude thatthe added peptide most likely exerts its effect through direct bind-ing to the MHC class I expressed on the target cell surface.

Peptide binding to MHC class I appears to alter or mask aninhibitory signal

NK recognition is thought to be mediated by an activating receptorwhose effects may then be overridden by an inhibitory receptor(1–3). Our results are most easily explained by assuming that bind-ing of the added peptide to the MHC class I on the target cellsurface altered or masked an inhibitory signal recognized by NKcells. However, in principle, the peptide binding to MHC class Imight either create a target structure that is recognized by an NK-activating receptor (i.e., similar to T cell recognition) or alter ormask an inhibitory structure that is recognized by an NK inhibitoryreceptor. In an attempt to distinguish between these possibilities,we performed cold target competition experiments. Radiolabeled,peptide-pulsed, normal B6 Con A lymphoblasts were incubatedwith B6 NK cells and varying numbers of cold B6 targets that hador had not been pulsed with peptide. The results (Fig. 2) show thatthe cold targets were equally effective competitors regardless ofwhether they were pulsed with OVAp peptide (Fig. 2A). This im-plies that both peptide-pulsed and unpulsed cold targets wereequally effective in forming conjugates with NK cells. We testedthis directly by measuring the ability of B6 NK cells to form con-jugates with B6 Con A lymphoblasts pulsed or not pulsed withOVAp and, as a control, with YAC-1 (Fig. 3). The NK cells werestained with a green fluorescent dye (FITC), and the Con A lym-phoblasts were stained with a red fluorescent dye (PKH26), mixed,and centrifuged together (experiment, conjugates should form) orkept suspended (control, conjugates much less likely to form).

Events detected in the flow cytometer that showed both green andred fluorescence were scored as conjugates. In agreement with thecompetition results (Fig. 2), comparable numbers of conjugatesformed using both pulsed and unpulsed Con A lymphoblasts (Fig.3). As only peptide-pulsed lymphoblasts are lysed, the observa-tions are consistent with the conclusion that binding of the addedpeptide to the MHC class I on Con A lymphoblast target cells isaltering or masking an inhibitory structure recognized by NK cells,leading to the lysis of target cells (seeDiscussion).

As a control for the cold target competition experiment, thesame cold target competition experiment was performed using anOVAp-specific CTL line as the effector cells (31). Here, as ex-pected, only OVAp-pulsed cold targets (and not Flu-NP-pulsed orunpulsed targets) were effective competitors (Fig. 2B), as it isknown that a specific peptide-MHC complex (H-b2m-pH) formsthe target structure that is recognized and activates the CTL to lysethe target cell (38). Similar cold target competition results for NKand CTL were obtained using the Kb-binding Flu-NP peptide (datanot shown).

Peptide-pulsed target cells remain sensitive to lysis mediated byCTL for a much longer period of time than by NK cells

To gain insight into the nature of the effect produced by peptidepulsing, the lifetime of the lysis-sensitive state was measured. Flu-NP-pulsed B6 Con A lymphoblasts were washed free of unboundpeptide and incubated for increasing lengths of time at 37°C beforeNK cells were added. The results show that the sensitivity of theseFlu-NP-pulsed target cells to NK-mediated lysis returned to that ofunpulsed targets following an incubation of 60 to 90 min (Fig. 4A).As a control, a CTL line specific for the same peptide was gener-ated (31) and tested against the same target cells pulsed with thesame peptide in the same assay. The peptide-pulsed targets re-tained full sensitivity to CTL-mediated lysis following up to 4 h ofincubation (Fig. 4B).

To verify that the short lifetime of the peptide-induced sensitivestate was not unique to Flu-NP peptide, we also tested the OVAppeptide in the same manner (Table I). Again, sensitivity to lysis ofOVAp-pulsed Con A lymphoblasts had returned to that of normal

FIGURE 1. Normal cells become more sensitive to NK lysis after being incubated with peptide that can bind to their MHC class I molecules.A, Percentlysis vs E:T ratio for target cells pulsed with Flu-NP (closed square), OVAp (closed diamond), or no peptide (open diamond). B6 Con A blasts were pulsedwith the peptide indicated (10 ng/ml) and then tested as targets at varying E:T cell ratios, as indicated on the abscissa, in a 4.5-h51Cr release cytotoxicityassay using syngeneic B6 NK cells. This experiment is representative of.40 such experiments.B, Peptide dose-response curve for Flu-NP (closed square),OVAp (closed diamond), or Flu-NP with GAD added on the N-terminus (open circle). B6 Con A blasts were pulsed with varying concentrations of peptideas indicated on the abscissa, washed free of unbound peptide, and then tested as targets in a 4.5-h51Cr release cytotoxicity assay using syngeneic B6 NKcells. The middle parts of both dose-response curves have been reproduced at least three times for both peptides using B6 NK cells derived from B6 normalmice, B6 CD8 knockout mice (23), and B6 athymic nude mice. This experiment is representative of two independent experiments.

758 CLASS I BINDING PEPTIDES AND NK RECOGNITION

by guest on Decem

ber 20, 2015http://w

ww

.jimm

unol.org/D

ownloaded from

Con A lymphoblasts after 90 min of incubation in the absence ofexogenous peptide. To test whether such targets could be resensi-tized to lysis, the same peptide (OVAp, 1 ng/well) was added toassay wells containing OVAp-pulsed Con A lymphoblasts incu-bated for 2 h in the absence of peptide (Table I, line 5); sensitivity

to NK-lysis was restored to that of targets tested immediately afterthe initial peptide pulsing (line 1). As a control, we also generatedan OVAp-specific CTL line and found that, as for Flu-NP, OVAp-pulsed target cells retained full sensitivity to lysis after 4 h ofpreincubation (data not shown).

FIGURE 2. Cold targets compete for NK-mediated lysis of hot targets regardless of whether they have been pulsed with peptide, whereas onlypeptide-pulsed cold targets can compete for CTL-mediated lysis.A, B6 radiolabeled Con A lymphoblasts, either pulsed (filled symbols) or unpulsed (opensymbols) with OVAp (1 ngm/ml), were tested as targets using B6 NK cells as in Figure 1B, except that unlabeled B6 Con A lymphoblasts, either pulsed(1 ng/ml; diamonds) or unpulsed (circles), were included at zero-, one-, three-, or fivefold multiplicities of the labeled targets, as indicated on the abscissa.Cold and hot targets were premixed before the addition of effector cells (i.e., NK cells). The figure is representative of six independent experiments.B, Sameas inA, except that an OVAp-specific CTL line was used at an effector to hot target ratio of 50:1. An additional group was included in which Flu-NP-pulsed(1 ng/ml) targets were tested as cold competitors (filled square). The figure is representative of four independent experiments.

FIGURE 3. B6 Con A lymphoblasts, whetherpulsed or not pulsed with peptide, are equally ef-fective in forming conjugates with B6 LAK cells.Targets (A, B6 Con A lymphoblasts;B, B6 Con Alymphoblasts pulsed with OVAp; orC, YAC-1cells (labeled with PKH26, detected by FL2; flu-orescence intensity is shown on they-axis)) werepelleted and incubated with effectors (B6 LAKcells (labeled with FITC, detected by FL1; fluo-rescence intensity is shown on thex-abscissa) atan E:T cell ratio of 3. Effector-target conjugateswere detected as PKH261FITC1 events, shownin the boxed area. As negative controls, FITC-labeled B6 LAK cells and PKH26-labeled targets(D, B6 Con A lymphoblasts;E, B6 Con A lym-phoblasts pulsed with OVAp;F, Yac-1 cells; atan E:T cell ratio of 3) were mixed and vortexedwithout any cocentrifugation before the analysiswith FACScan. Fewer effector-target conjugates(PKH261FITC1 events) were formed when theeffectors and targets were not brought together bycentrifugation. Note that 10,000 events/samplewere analyzed, but only the first 2,000 eventswere presented in the plot for the clarity and neat-ness of the presentation. The figure is represen-tative of three independent experiments.

759The Journal of Immunology

by guest on Decem

ber 20, 2015http://w

ww

.jimm

unol.org/D

ownloaded from

BFA prevents the loss of sensitivity to NK-mediated lysis

It is possible that the expression of newly synthesized MHC classI molecules is involved in the loss of sensitivity to NK lysis ofpeptide-pulsed target cells after the 90 min of preincubation. BFAis a fungal metabolite that reversibly disrupts the Golgi apparatus,resulting in the blocking of transport to the cell surface of newlysynthesized protein (39). In particular, BFA has been shown toblock the transport of MHC class I molecules to the cell surface(40). We tested the effect of including BFA in the preincubationstep of the experiments shown in Figure 4; Flu-NP-pulsed B6 ConA lymphoblasts were incubated in the absence of free exogenouspeptide with or without BFA for varying lengths of time beforeNK cells or CTL were added. In the presence of BFA, the sensi-tivity of the Flu-NP-pulsed target cells to NK-mediated lysis re-mained high for at least 2 h instead of rapidly falling (Fig. 5A).CTL-mediated lysis of Flu-NP pulsed target cells was not affectedin the presence of BFA (Fig. 5B). Furthermore, background lysisof normal Con A lymphoblasts was not affected by BFA; BFA didnot sensitize normal cells to NK lysis in the absence of peptide.

The presence of BFA also prevented the loss of sensitivity to NKlysis for OVAp-pulsed Con A lymphoblasts (Fig. 5C). Thus, weconclude that preventing the appearance of newly synthesized pro-teins, most likely MHC class I, on the cell surface prevents the lossof sensitivity to NK lysis of peptide-pulsed target cells. The pos-sibility that BFA is having some other effect on MHC class I ex-pression is explicitly addressed in the following section.

BFA prevents the appearance of newly synthesized empty MHCclass I but has little effect on overall MHC class I expression

One explanation for the peptide-induced sensitization to NK lysisis that NK inhibitory receptors recognize empty MHC class I mol-ecules on the target cell surface and that the addition of high af-finity peptide fills the empty MHC class I molecules. Therefore,cells become sensitive to NK lysis because the added peptideblocks the NK recognition of inhibitory ligand. In the absence ofexogenous peptide, newly synthesized empty MHC class I mole-cules emerge onto the target cell surface and regenerate the inhib-itory signal, thus preventing lysis.

To test directly for a correlation between the absence of emptyMHC class I and sensitivity to NK lysis, and the reappearance ofempty MHC class I and the loss of sensitivity to lysis, we mea-sured the relative number of full and empty Kb molecules on ConA lymphoblasts pulsed with OVAp (Kb-specific). OVAp-pulsedCon A lymphoblasts were washed free of unbound OVAp andincubated at 37°C for 90 min in the presence or the absence ofBFA. The expression of peptide-Kb complex was measured beforeand after the 90-min incubation using the mAb, 5F1, which rec-ognizes specifically the trimolecular Kb complex (H-b2m-p) (34,41). We found that the level of peptide-Kb complex expression wasnot affected by the 90-min incubation in the absence of BFA (Fig.6, a andb), but fell slightly (;10–20%) in the presence of BFA(Fig. 6c), perhaps because BFA prevents the transport of newlysynthesized trimolecular MHC class I to the cell surface whilehaving no effect on the endocytosis of cell surface proteins. EmptyKb molecules appeared only in the absence of BFA (Fig. 6e). Todetect empty Kb molecules, we used an OVAp peptide in which

FIGURE 4. The NK target structure formed by pulsing with peptide is short lived compared with the CTL target structure formed by pulsing the sametargets with the same peptide.A, Time delay experiment, NK cells. B6 Con A lymphoblasts were pulsed with 1 ng/ml Flu-NP (filled square), a nonbindingpeptide, Tum2 (filled circle), or no peptide (open diamond) and washed free of unbound peptide as described in Figure 1. The cells were then incubatedin CM at 37°C for varying lengths of time, as indicated on the abscissa, before being tested as targets in a 4.5-h51Cr release assay using syngeneic B6NK cells as described in Figure 1. The figure is representative of eight independent experiments.B, Time delay experiment, CTL. B6 Con A lymphoblasttargets were pulsed with Flu-NP (filled square) or no peptide (open diamond), washed free of unbound peptide, and incubated for varying lengths of timein CM exactly as described in A. They were then tested for their ability to be lysed by a CTL line specific for the peptide using conditions identical withthose used for the NK cytotoxicity assay, except that the E:T cell ratio was 10:1. The figure is representative of three independent experiments.

Table I. Time delay experiment for NK cells using OVAp peptidea

Time Delay (h) % Specific Lysis

1 0 55.96 4.0 (21.66 2.3)2 1 41.86 2.4 (22.36 3.5)3 1.5 22.86 3.0 (24.26 2.5)4 2 29.26 3.8 (20.66 1.9)5 2 48.26 5.1b

a B6 Con A lymphoblasts were pulsed with 1 ngm/ml OVAp and washed free ofunbound peptide, as in Figure 1. They were then incubated in medium at 37°C forvarious lengths of time as indicated under Time Delay before being used as targets ina 4.5-h51Cr release assay, as in Figure 1. The entry in parentheses after each %Specific Lysis entry is the background lysis observed using targets that were notpulsed with peptide before the time delay but were otherwise treated identically. Aftera 2-h time delay step, one set of target cells was cocultured with NK cells in thepresence of added peptide (1 ng/well) during the51Cr release assay (line 5). Data fromone of two identical experiments are shown.

b Same as line 4 except 1 ng of OVAp peptide was added to each well at the startof the 51Cr release assay.

760 CLASS I BINDING PEPTIDES AND NK RECOGNITION

by guest on Decem

ber 20, 2015http://w

ww

.jimm

unol.org/D

ownloaded from

the lysine (K) at position 7 was biotinylated (OVApK-bio). Thislysine side chain is known to be one of the CTL epitopes in OVApand is therefore expected to protrude from the peptide bindinggroove (42, 43). We found that this peptide binds specifically to Kb

and can be readily detected by the addition of streptavidin-phyco-erythrin (Fig. 6 and data not shown). As a control peptide, we usedOVAp to which a biotinylated aminocaproic acid was added to theN-terminus (OVApX-bio). This peptide did not bind (Fig. 6 anddata not shown).

Newly expressed empty Kb molecules were clearly detectableon pulsing OVAp-pulsed lymphoblasts with OVApK-bio after the90-min incubation in the absence of exogenous peptide (Fig. 6e)but were not detectable when BFA was present during the 90-minincubation (Fig. 6f ). The total number of OVApK-bio-bound Kb

complexes was also measured and found to decline after the 90-min incubation (;55% in the absence of BFA and 53% in thepresence of BFA; staining data not shown); nevertheless, the im-portant point was that the decline was not affected by the presenceof BFA. The large decline might be a result of the OVApK-bio

peptide having a greatly reduced binding affinity to class I as aresult of the modification. This appears to be the case. Approxi-mately a 10-fold higher concentration of OVApK-bio was requiredto stabilize Kb molecules on the cell surface of RMA-S cells com-pared with that of OVAp (data not shown).

Clearly, the loss of empty Kb molecules after peptide pulsingcorrelated with the sensitivity of these target cells to NK lysis, andthe reappearance of empty Kb molecules after the 90-min incuba-tion at 37°C coincides with the loss of sensitivity to NK lysis.Thus, these data are fully consistent with our hypothesis that NKinhibitory receptors recognize empty class I molecules.

A potential problem with this model is the observation (Fig. 5)that Con A lymphoblasts not pulsed with peptide and incubatedwith BFA remained resistant to lysis. Empty MHC class I mole-cules are known to be unstable, and if BFA blocks the expressionof new empty MHC class I, one might expect all empty MHC classI to disappear over the time of the assay, thus rendering the Con Alymphoblasts sensitive to lysis. To address this possibility, wemeasured the relative number of empty Kb molecules on Con Alymphoblasts not pulsed with peptide and incubated with or with-out BFA for 0 and 4 h using OVApK-bio as shown in Figure 6. Therelative number of empty Kb fell 31 6 7% in the presence of BFA

and 116 12% in the absence of BFA over the 4-h incubation. Wehypothesize that the remaining empty MHC class I molecules aresufficient to provide protection from NK lysis (seeDiscussion).

The Ly49A molecule does not recognize empty Dd molecules

In contrast to our observations, other groups have shown that underappropriate conditions, addition of H-2Dd specific peptide createsan inhibitory signal that protects NK-susceptible target cells ex-pressing H-2Dd from being lysed by Ly49A1 B6 NK cells (12,13). To attempt to reconcile this difference, we studied the recog-nition of Dd molecules by Ly49A1 and Ly49A2 NK cells in oursyngeneic experimental system. The Con A lymphoblasts and NKcells used were derived, respectively, from normal and athymicnude BALB/c (H-2d) mice. Day 3, rIL-2-activated NK cells weresorted into Ly49A1 and Ly49A2 subsets using the mAb JR9-318,which recognizes Ly49A molecules on both B6 and BALB/c NKcells (35, 44). BALB/c Con A lymphoblasts, both pulsed and notpulsed with a Dd-specific peptide, HIVgp160318–327(HIVp) (27),were examined for sensitivity to lysis mediated by either Ly49A1

NK cells or Ly49A2 NK cells. The results show that BALB/clymphoblasts, whether pulsed or not with HIVp, were resistant tolysis mediated by the Ly49A1 NK cells, but when F(ab9)2 anti-Ly49A mAb (JR9-318) was included in the assay, both normal andHIVp-pulsed lymphoblasts were lysed by Ly49A1 NK cells (Fig.7A; seeDiscussion). In contrast, when Ly49A2 NK cells wereused, they lysed HIVp-pulsed lymphoblasts and spared unpulsedlymphoblasts regardless of whether F(ab9)2 anti-Ly49A mAb waspresent (Fig. 7B). When a Kd-specific peptide (Flu-NP-Kd) wasused for pulsing BALB/c lymphoblasts, both Ly49A1 andLy49A2 NK populations could produce lysis. Interestingly, a mix-ture of both Flu-NP-Kd and HIVp peptide in the absence of F(ab9)2anti-Ly49A mAb (JR9-318) enabled lysis by Ly49A1 as well asLy49A2 NK cells (Fig. 7,C andD; seeDiscussion).

The observation that HIVp-pulsed targets were not lysed byLy49A1 NK cells is fully consistent with the hypothesis thatLy49A recognizes peptide-loaded Dd molecules and prevents lysisthat would otherwise have occurred. That these targets were lysedwhen Ly49A molecules on the NK cells were covered up byF(ab9)2 anti-Ly49A mAb and were also lysed when Ly49A2 NKcells were used as NK effectors is consistent with the existence of

FIGURE 5. The NK target structure formed by incubation with peptide is stable in the presence of BFA.A, B6 Con A lymphoblasts were pulsed at 4°Cwith 1 ng/ml Flu-NP (squares) or no peptide (triangles), washed free of peptide, and incubated in CM at 37°C for varying lengths of time (abscissa) asdescribed in Figure 3Awith (filled symbols) or without (open symbols) added BFA (5mg/ml) before being used as targets for B6 NK cells as describedin Figure 1. BFA (0.4mg/ml) was also included in the cytotoxicity assay for those groups (filled symbols) for which it had been used previously. The figureis representative of six independent experiments.B, Same as inA except that the effector cells were Flu-NP specific CTL generated and analyzed asdescribed in Figure 3B. The figure is representative of four independent experiments.C, Same as inA except that OVAp (1 ng/ml; squares) was used topulse the B6 Con A lymphoblasts. The figure is representative of two independent experiments.

761The Journal of Immunology

by guest on Decem

ber 20, 2015http://w

ww

.jimm

unol.org/D

ownloaded from

inhibitory receptors recognizing empty MHC class I, as hypothe-sized in the preceding sections. An explicit model that is consistentwith all the lysis results in Figure 7, in which resistance to lysisdepends upon the total number of possible inhibitory signals, isgiven in Discussion(Table II).

DiscussionSelf-recognition involving empty MHC class I molecules

We here confirm previous results (8, 9) that normal Con A-inducedlymphoblasts become sensitive to lysis by syngeneic NK cellswhen incubated with peptide that can bind to their MHC class Imolecules (Fig. 1). The process does not seem to be peptide se-quence specific, in that all peptides tested sensitized targets for NKlysis provided that they could bind to MHC class I in the peptide-pulsing procedure. Concentrations of 1 to 10 pg/ml of the peptidetested (all of which bind with high affinity) were sufficient to pro-duce significant sensitization (Fig. 1). For comparison, concentra-tions 100-fold lower are sufficient to sensitize lymphoblasts to lysisby CTL lines specific for the same or similar high affinitypeptides (36).

It is now widely accepted that if an activating receptor on an NKcell recognizes a cell, that cell will be killed unless an inhibitory

receptor on the NK cell also recognizes the cell. According to thismodel, the added peptide in our system must be altering the targetcell either by creating a new ligand recognized by an activatingreceptor or by altering or masking a ligand recognized by an in-hibitory receptor (or possibly both). To distinguish between thesetwo possibilities, the most powerful approach is to use specificmAb F(ab9)2 fragments against the ligand as has been done toblock the inhibitory MHC class I interaction (10, 45). However,this approach cannot yet be used in this study because the putativereceptor involved in our system has yet to be identified (except thatLy49A is not involved). We have relied on cold target competitionas an alternative for providing insight into the nature of the ligandaffected by peptide pulsing and found that unlabeled Con A lym-phoblast targets, regardless of whether peptide pulsed, wereequally effective competitors for NK-mediated lysis of labeledpeptide-pulsed targets (Fig. 2). We also found, using flow cytom-etry, that lymphoblasts pulsed or not pulsed with peptide wereequally effective in forming conjugates with FITC-labeled NKcells (Fig. 3). Similarly, Ljunggren et al. found that the cell lineRMA, which is moderately resistant to NK lysis, and RMA-S, amutant cell line derived from it and highly sensitive to NK lysis,were equivalent in the ability to bind NK cells (46). Taken together

FIGURE 6. Total peptide-Kb expression was lit-tle affected by BFA, but newly synthesized emptyKb molecules emerged only in the absence of BFAduring a 90-min incubation.A, FACS profiles forpeptide-Kb expression (left panels) and for emptyKb expression (right panels). The left panels(a–c)show peptide-Kb complex expression (measuredwith mAb 5F1 staining) immediately after peptidepulsing (a) and 90 min later in the absence (b) or thepresence (c) of BFA. The right panels(d–f) showbinding of biotinylated OVAp immediately afterpeptide pulsing (d) and 90 min later in the absence(e) or the presence (f) of BFA. OVApX-bio, in whichX is an aminocaproic acid serving as a linker be-tween biotin and the peptide (does not bind to MHCclass I), and OVApK-bio, which binds to Kb, wereused in the staining assay. The binding of biotinyl-ated OVAp was visualized with R-phycoerythrin-conjugated streptavidin (Sigma), which binds to bi-otin, and were analyzed using LYSIS II program(Becton Dickinson). The figure is representative ofsix independent experiments.B is the mean channelvalue 6 SD of three replicates of the nine profilesshown in theright panels(d–f).

762 CLASS I BINDING PEPTIDES AND NK RECOGNITION

by guest on Decem

ber 20, 2015http://w

ww

.jimm

unol.org/D

ownloaded from

with the fact that only peptide-pulsed targets are lysed, we con-clude that the added peptide is most likely altering or maskingthe ligand recognized by an NK inhibitory receptor. The BFAexperiments support this conclusion. Preventing surface arrivalof proteins should not affect an activating ligand that is alreadythere.

The observation that the ligand recognized by the NK inhibitoryreceptor operative in this system and the CTL receptor are affected

differently by BFA leads to the hypothesis that they are recogniz-ing different ligands and, in particular, that the inhibitory NK re-ceptor recognizes the empty form of MHC class I molecules onsyngeneic lymphoblasts. The data presented in Figures 1 to 5 canbe explained by and support this hypothesis. Thus, normal lym-phoblasts can be recognized by syngeneic NK cells, but their lysisis normally prevented when negative inhibitory signals generatedby recognition of empty MHC class I molecules are above some

FIGURE 7. Ly49A1 NK cells lyse Flu-NP (Kd-specific)-pulsed Con A lymphoblasts, but not HIVp (Dd-specific)-pulsed Con A lymphoblasts.A,BALB/c Con A lymphoblasts, whether pulsed with HIVp (filled square) or with CM alone (open square), were resistant to lysis mediated by Ly49A1 NKcells. The addition of F(ab9)2 anti-Ly49A mAb (JR9-318) in the assay made both HIVp-pulsed lymphoblasts (closed triangles) and CM-pulsed lymphoblasts(open triangles) susceptible to lysis by Ly49A1 NK cells. The figure is representative of three independent experiments.B, Normal lymphoblasts pulsedwith CM alone (open symbols) were resistant to lysis by Ly49A2 NK cells in either the presence (triangles) or the absence (squares) of F(ab9)2 anti-Ly49AmAb (JR9-318), while HIVp-pulsed lymphoblasts (closed symbols) were lysed by Ly49A2 NK cells regardless of whether the F(ab9)2 anti-Ly49A mAb(JR9-318) was added.C, Both CM-pulsed (open circle) and HIVp-pulsed (filled squares) lymphoblasts were resistant to lysis by Ly49A1 NK cells, whilelymphoblasts pulsed either with Flu-NP-Kd peptide (Kd-specific, closed circles) alone or with a mixture of both Flu-NP peptide and HIVp (closed diamonds)were lysed.D, Lymphoblasts pulsed with Flu-NP-Kd (closed circles), HIVp (closed squares), or both (closed diamonds) were lysed by Ly49A-NK cells,while normal lymphoblasts (open circles) remained resistant to lysis. The figure is representative of three independent experiments.

Table II. Model for recognition by Ly49A1 NK cells including two additional inhibitory receptors

Target CellsLysis Seenin Figure 6 Dd(f)a Dd(e)b Kd(e)c

Summation ofInhibitory Signals

Normal lymphoblast No 0.9d 0.1 0.1 1.1Lymphoblast1 Dd-peptide No 1.0 0 0.1 1.1Lymphoblast1 Kd-peptide Yes 0.9 0.1 0 1.0Lymphoblast1 Kd-peptide1Dd-peptide Yes 1.0 0 0 1.0Lymphoblast1 JR9-318 F(ab9)2 Yes 0 0.1 0.1 0.2Lymphoblast1 Dd-peptide1 JR9-318 F(ab9)2 Yes 0 0 0.1 0.1Lymphoblast1 Kd-peptide1 JR9-318 F(ab9)2 Yes 0 0.1 0 0.1

a Dd(f), inhibitory signal from Ly49A receptor recognizing peptide-bound Dd.b Dd(e), inhibitory signal from receptors recognizing empty Dd.c Kd(e), inhibitory signal from receptors recognizing empty Kd.d The numbers are the inhibitory signal strengths assigned to each receptor involved in NK recognition.

763The Journal of Immunology

by guest on Decem

ber 20, 2015http://w

ww

.jimm

unol.org/D

ownloaded from

threshold level and override the activation signal in the NK cell.Pulsing normal lymphoblasts with peptide of high binding affinityfills most (if not all) empty MHC class I molecules and thus re-duces the negative inhibitory signal below the threshold level insome NK cells and renders the lymphoblasts susceptible to lysis(Fig. 1). When peptide-pulsed lymphoblasts are incubated in theabsence of exogenous peptide (preincubation experiment, Fig.3A), newly synthesized empty MHC class I molecules are trans-ported to and expressed on the cell surface, where they regeneratethe inhibitory structure that increases the inhibitory signal abovethe threshold level and thus prevents NK lysis. Furthermore, ifMHC class I-specific peptide is added to the target cells again,lysis can be restored (Table I). If the “regeneration” of the inhib-itory structure is prevented by blocking the expression of newlysynthesized MHC class I in the preincubation step (BFA experi-ment, Fig. 4), the targets remain sensitive to lysis.

An understanding of the number of “empty” MHC class I mol-ecules on the cell surface under varying conditions is central to ourmodel. New empty MHC class I molecules might arise on the cellsurface through loss of peptide from the trimolecular complex onthe cell surface, as both peptide andb2m can freely and indepen-dently disassociate from the trimolecular complex (19, 47, 48).Alternatively, newly synthesized empty MHC class I moleculescan also arrive at the cell surface (14, 15). Whether MHC class Imolecules are truly empty or contain peptide binding with lowaffinity (pL) that is readily lost is not clear. To distinguish betweenthese possibilities, we directly examined the relative frequency offull and empty Kb MHC class I on B6 Con A lymphoblasts pulsedwith a high affinity binding peptide (OVAp) and then incubated for90 min in the presence or the absence of BFA (Fig. 6). In thepresence of BFA, which should prevent the emergence of newempties from the cell interior, no new empties were detected aftera 90-min incubation, implying that the trimolecular complex has ahalf-life much greater than 90 min and that little peptide was lostduring this 90-min incubation. There is no direct measurement ofthe half-life of this trimolecular complex. However, a possiblysimilar Flu-NP-Db trimolecular complex (the same as studied here,see Fig. 1) is known to have a half-life of 10 h (48). In the absenceof BFA, a significant number of new empties appeared on the cellsurface (Fig. 6). We assume that these are newly synthesized mol-ecules exported from the cell interior. Their appearance correlateswith the disappearance of sensitivity to NK lysis (Figs. 4 and 5). Inexamining the effect of BFA on normal B6 lymphoblasts, wefound that empty MHC class I were still detected even after thecells were incubated for 4 h in the presence of BFA at 37°C,although a decrease in the level of empty MHC class I expressionwas observed. Given that the H-b2m forms of Db and Kb mole-cules have been reported to have half-lives much less than 2 h (15,19, 20), one might have expected unpulsed Con A lymphoblasts tohave lost all their empty MHC class I molecules during the incu-bation with BFA. That they did not has two possible explanations:1) new empty MHC class I molecules are continuously formedthrough loss of low affinity peptide from trimolecular MHC classI complexes (pL-H-b2m) already on the cell surface; and 2) mea-surements of H-b2m half-lives have been made by extractingH-b2m complexes from the cell surface with mAb. Molecules em-bedded in the membrane of a normal, viable cell may be morestable. We conclude that the ligand for the inhibitory receptor op-erative in our system is most likely to be empty MHC class I.

The model that added peptides might be displacing protective“self” peptides is rendered unlikely by the current data

A 45-min pulse with a high affinity peptide produced a state ofsensitization (Fig. 1). If the high affinity peptide is displacing par-

ticular protective self peptides, then they must be bound with lowaffinity to be displaced in such a short time pulse (19). This, then,implies that control target cells not pulsed with high affinitypeptide and then incubated with BFA should have become sen-sitive to lysis as the protective self peptide was lost. This wasnot seen (Fig. 5).

An as yet unidentified inhibitory receptor, differing from Ly49A,recognizes the bimolecular form of the Dd molecule

As described in the introduction, three groups have shown that,under appropriate conditions, addition of MHC class I bindingpeptides to a target can prevent NK lysis (12, 13, 16). To reconcilethe difference between these published and our experimental data,we studied the recognition of the Dd molecule by Ly49A1 andLy49A2 NK subsets in our syngeneic experimental system. Wefound that the Ly49A1 subset of NK cells could not lyse syngeneicDd-bearing lymphoblasts pulsed with Dd-binding peptide (Fig. 7),consistent with the Ly49A inhibitory receptor recognizing the Dd

trimolecular complex and providing a dominant negative signal(Fig. 7). In agreement with this, the targets were lysed whenLy49A was covered up by F(ab9)2 anti-Ly49A mAb. The Ly49A2

subset of NK cells killed the same syngeneic Dd-bearing lympho-blasts pulsed with Dd-binding peptide, consistent with our hypoth-esis that there might be an as yet unidentified inhibitory receptor(which may or may not be a member of the Ly49 family) thatrecognizes the empty form of the Dd molecule.

In support of our hypothesis that there are inhibitory receptorsrecognizing empty MHC class I molecules, a recent report (49)concluded that cell surface expression of human MHC class I mol-ecules, in the absence of peptide, was both necessary and sufficientto inhibit HLA-specific human NK lines and clones. They trans-fected RMA-S cells with human HLA-C of two different allotypesalong with humanb2m. Culture of the cells at 26°C without ex-ogenous peptide allowed for high expression of the transfectedclass I, and this persisted for at least 2 h after the cells were trans-ferred to 37°C. The presence of a particular empty HLA-C allotypewas sufficient to inhibit lysis by an NK clone specifically inhibitedby that allotype. Note that the inhibitory receptors involved in thisstudy are most likely the members of the NK inhibitory receptorfamily, structurally unrelated to the Ly49 family (50, 51).

A teeter-totter model for resistance vs sensitivity to NK lysis

Correa et al. have shown that the Ly49A-Dd interaction is suffi-cient to inhibit all types of NK cell activation pathways that havebeen examined, but the contrary has been observed in this study(52). Our data showed that Ly49A1 NK cells could lyse Dd-ex-pressing lymphoblasts pulsed with Flu-NP-Kd peptide even if theywere also pulsed with Dd-binding peptide (Fig. 6C). This apparentdiscrepancy can be explained by the following model. 1) Individ-ual NK cells have different inhibitory receptors that can recognizeeither empty or full MHC class I molecules. 2) The strength of theinhibitory signal generated by a particular receptor is proportionalto the number of MHC class I molecules it can recognize. 3) Forinhibition of lysis to occur, the summation of all inhibitory signalsmust exceed some critical threshold value.

Let us apply this model to all the data in Figure 7 using Ly49A1

NK cells (Table II). We assume that 10% (0.1) of Kd and Dd

molecules on the lymphoblasts used are empty, as has been re-ported (14) for Db MHC class I molecules, but would reach thesame conclusions for any value.0 and,1. For normal lympho-blasts (Table II, line 1), there is a total inhibitory signal of 1.1 (0.9(from Ly49A recognizing peptide-bound Dd) plus 0.1 (from a new

764 CLASS I BINDING PEPTIDES AND NK RECOGNITION

by guest on Decem

ber 20, 2015http://w

ww

.jimm

unol.org/D

ownloaded from

receptor recognizing empty Dd) plus 0.1 (from a second new re-ceptor recognizing empty Kd); no lysis is seen. When the lympho-blasts are pulsed with Dd-specific peptide (Table II, line 2), thetotal inhibitory signal remains 1.1, because as the Ly49A signalgoes up by 0.1, the inhibitory signal generated by the receptorrecognizing empty Dd goes down by the same amount, 0.1, andagain no lysis is seen. However, when they were pulsed with Kd-specific peptide (line 3) or with both Kd- and Dd- specific peptide(line 4), the total inhibitory signal falls to 1.0; lysis is now seen. Ingoing down Table II, one sees that lysis was observed wheneverthe summation of inhibitory signals was 1.0 or less. Comparison oflines 2 and 4 is particularly interesting, in that pulsing lympho-blasts with Dd-binding peptide alone does not block inhibition(line 2), but pulsing lymphoblasts with both Kd- and Dd-bindingpeptide does (line 4).

The model implies that there is a critical balancing of activatingand inhibitory signals leading either to sensitivity or resistance tolysis. It is much like a teeter-totter in a children’s playground, inwhich a given end is either fully up or fully down depending uponthe balance of the forces last acting on the two ends. Whether thereis a subset of B6 NK cells with an inhibitory receptor that recog-nizes peptide-bound Kd molecules cannot be determined fromthese data, as inhibition or activation of such a subset is difficult todetect unless the subset is relatively pure. We could detect theeffect of the Ly49A inhibitory receptor in our system only afterpurifying Ly49A1 cells.

Most previous studies (for an exception, see Ref. 53) supportingthe existence of negative-signaling NK receptors have involved theprotection from lysis of allogeneic target cells recognized by in-hibitory receptors. The data presented here provide direct evidencethat negative-signaling receptors can also protect normal synge-neic target cells from lysis. They also suggest a possible explana-tion for why some virus-infected cells become targets for synge-neic NK cells: as a result of the virus infection, very few emptyMHC class I molecules are exported to the cell surface either be-cause very large quantities of viral peptide inside the cell saturateMHC class I or because the virus greatly reduces overall MHCclass I production such that few empties (albeit a higher percentageof all class I) reach the cell surface.

AcknowledgmentsWe thank Dr. D. Raulet (University of California, Berkeley) for the kindgift of hybridoma JR9-318 with permission from Dr. J. Roland (PasteurInstitute, Paris, France), and J. Ferguson for help in HPLC purification ofpeptides and synthesizing the biotinylated peptides.

References1. Moretta, L., E. Ciccone, A. Moretta, P. Hoglund, C. Ohlen, and K. Karre. 1992.

Allorecognition by NK cells: nonself or self.Immunol. Today 13:300.2. Yokoyama, W. M., and W. E. Seaman. 1993. The Ly-49 and NKR-P1 gene

families encoding lectin-like receptors on natural killer cells: the NK gene com-plex. Annu. Rev. Immunol. 11:613.

3. Lanier, L. L., and J. H. Phillips. 1996. Inhibitory MHC class I receptors on NKcells and T cells.Immunol. Today 17:86.

4. Storkus, W. J., D. N. Howell, R. D. Salter, J. R. Dawson, and P. Cresswell. 1987.NK susceptibility varies inversely with target cell class I HLA antigen expres-sion.J. Immunol. 138:1657.

5. Storkus, W. J., J. Alexander, J. A. Payne, P. Cresswell, and J. R. Dawson. 1989.Thea1/a2 domains of class I HLA molecules confer resistance to natural killing.J. Immunol. 143:3853.

6. Storkus, W. J., R. D. Salter, J. Alexander, F. E. Ward, R. E. Ruiz, P. Cresswell,and J. R. Dawson. 1991. Class I-induced resistance to natural killing: identifica-tion of nonpermissive residues in HLA-A2.Proc. Natl. Acad. Sci USA 88:5989.

7. Storkus, W. J., R. D. Salter, P. Cresswell, and J. R. Dawson. 1992. Peptide-induced modulation of target cell sensitivity to natural killing.J. Immunol. 149:1185.

8. Chadwick, B. S., and R. G. Miller. 1992. Hybrid resistance in vitro: possible roleof both class I MHC and self peptides in determining the level of target cellsensitivity.J. Immunol. 148:2307.

9. Chadwick, B. S., S. R. Sambhara, Y. Sasakura, and R. G. Miller. 1992. Effect ofclass I MHC binding peptides on recognition by natural killer cells.J. Immunol.149:3150.

10. Karlhofer, F. M., R. K. Ribaudo, and W. M. Yokoyama. 1992. MHC class Ialloantigen specificity of Ly49A1 IL-2-activated natural killer cells.Nature 358:66.

11. Kane, K. P. 1994. Ly49 mediates EL4 lymphoma adhesion to isolated class Imajor histocompatibility complex molecules.J. Exp. Med. 179:1011.

12. Correa, L., and D. H. Raulet. 1995. Binding of diverse peptides to MHC class Imolecules inhibits target cell lysis by activated natural killer cells.Immunity 2:61.

13. Orihuela, M., D. H. Margulies, and W. M. Yokoyama. 1996. The natural killercell receptor Ly49A recognizes a peptide-induced conformational determinant onits major histocompatibility complex class I ligand.Proc. Natl. Acad. Sci. USA93:11792.

14. Christinck, E. R., M. A. Luscher, B. H. Barber, and D. B. Williams. 1991. Peptidebinding to class I MHC on living cells and quantitation of complexes required forCTL lysis. Nature 352:67.

15. Ortiz-Navarrete, V., and H. G. J. 1991. Surface appearance and instability ofempty H-2 class I molecules under physiological conditions.Proc. Natl. Acad.Sci. USA 88:3594.

16. Malnati, M. S., M. Peruzzi, K. C. Parker, W. E. Biddison, E. Ciccone, A. Moretta,and E. O. Long. 1995. Peptide specificity in the recognition of MHC class I bynatural killer cell clones.Science 267:1016.

17. Peruzzi, M., K. C. Parker, E. O. Long, and M. S. Malnati. 1996. Peptide sequencerequirements for the recognition of HLA-B*2705 by specific natural killer cells.J. Immunol. 157:3350.

18. Long, E. O., D. N. Burshtyn, W. P. Clark, M. Peruzzi, S. Rajagopalan, S. Rojo,N. Wagtman, and C. C. Winter. 1997. Killer cell inhibitory receptors: diversityspecificity and function.Immunol. Rev. 155:135.

19. Neefjes, J. J., L. Smith, M. Gehrmann, and H. L. Ploegh. 1992. The fate of thethree subunits of major histocompatibility complex class I molecules.Eur. J. Im-munol. 22:1609.

20. Burshtyn, D. N., and B. H. Barber. 1993. Dynamics of peptide binding to purifiedantibody-bound H-2Db and H-2Db-b2 m. J. Immunol. 151:3082.

21. Bjorkman, P. J., and P. Parham. 1990. Structure, function, and diversity of classI major histocompatibility molecules.Annu. Rev. Biochem. 59:253.

22. Kung, S. K. P., and R. G. Miller. 1995. The NK1.1 antigen in NK-mediated F1antiparent killer in vitro.J. Immunol. 154:1624.

23. Fung-Leung, W.-P., M. W. Schilham, A. Rahemtulla, T. M. Kundig,M. Vollenweider, J. Potter, W. van Ewijk, and T. W. Mak. 1991. CD8 is neededfor development of cytotoxic T cells but not helper T cells.Cell 65:443.

24. Karasuyama, H., H. Tohyama, and T. Tada. 1989. Autocrine growth and tumor-igenicity of interleukin 2-dependent helper T cells transfected with IL-2 gene.J. Exp. Med. 169:13.

25. Rotzschke, O., K. Falk, K. Deres, H. Schild, M. Norda, J. Metzger, G. Jung, andH.-G. Rammensee. 1990. Isolation and analysis of naturally processed viral pep-tides as recognized by cytotoxic T cells.Nature 348:252.

26. Rotzschke, O., K. Falk, S. Stevanovie, G. Jung, P. Walden, and H.-G. Ram-mensee. 1991. Exact prediction of a natural T cell epitope.Eur. J. Immunol.21:2891.

27. Shiral, M., C. D. Pendleton, and J. A. Berzofsky. 1992. Broad recognition ofcytotoxic T cell epitopes from the HIV-1 envelope protein with multiple class Ihistocompatibility molecules.J. Immunol. 148:1657.

28. Romero, P., G. Corradin, I. F. Luescher, and J. L. Maryanski. 1991. H-2Kd-restricted antigenic peptides share a simple binding motif.J. Exp. Med. 174:603.

29. Lurquin, C., A. Van Pel, B. Mariame, E. De Plaen, J.-P. Szikora, C. Janssens,M. J. Reddehase, J. Lejeune, and T. Boon. 1989. Structure of the gene of Tum2

transplantation antigen P91A: the mutated exon encodes a peptide recognizedwith Ld by cytotoxic T cells.Cell 58:293.

30. Kozlowski, S., M. Corr, T. Takeshita, L. F. Boyd, C. D. Pendleton,R. N. Germain, J. A. Berzofsky, and D. H. Margulies. 1992. Serum angiotensin-1converting enzyme activity processes a human immunodeficiency virus 1 gp 160peptide for presentation by major histocompatibility complex class I molecules.J. Exp. Med. 175:1417.

31. Sambhara, S. R., A. G. Upadhya, and R. G. Miller. 1990. Generation and char-acterization of peptide-specific, MHC-restricted cytotoxic T lymphocyte (CTL)and helper T cell lines from unprimed T cells under microculture conditions.J. Immunol. Methods 130:101.

32. Kung, K. P., and R. G. Miller. 1997. Mouse natural killer subsets defined by theirtarget specificity and the ability to be rendered unresponsive separately in vivo.J. Immunol. 158:2616.

33. Cavarec, L., A. Quillet-Mary, D. Fradelizi, and H. Conjeaud. 1990. An improveddouble fluorescence flow cytometry method for the quantification of killer cell/target cell conjugate formation.J. Immunol. Methods 130:251.

34. Sherman, L. A., and C. P. Randolph. 1981. Monoclonal anti-H-2Kb antibodiesdetect serological differences between H-2Kb mutants.Immunogenetics 12:183.

35. Roland, J., and P. A. Cazenave. 1992. Ly-49 antigen defined anab TCR popu-lation in i-IEL with an extrathymic maturation.Int. Immunol. 4:699.

36. Rammensee, H.-G., K. Falk, and O. Rotzschke. 1993. Peptides naturally pre-sented by MHC class I molecules.Annu. Rev. Immunol. 11:213.

37. Karre, K., H.-G. Ljunggren, B. Piontek, and R. Kiessling. 1986. Selective rejec-tion of H-2-deficient lymphoma variants suggests alternative immune defencestrategy.Nature 319:675.

38. Townsend, A., and H. Bodmer. 1989. Antigen recognition by class I-restricted Tlymphocytes.Annu. Rev. Immunol. 7:601.

765The Journal of Immunology

by guest on Decem

ber 20, 2015http://w

ww

.jimm

unol.org/D

ownloaded from

39. Misumi, Y., Y. Misumi, K. Miki, A. Takatsudi, G. Tamura, and Y. Ikehara. 1986.Novel blockade by Brefeldin A of intracellular transport of secretory proteins incultured rat hepatocytes.J. Biol. Chem. 261:11398.

40. Yewdell, J. W., and J. R. Bennink. 1992. Brefeldin A specifically inhibits pre-sentation of protein antigens to cytotoxic T lymphocytes.Science 244:1071.

41. Bluestone, J. A., S. Jameson, S. Miller, and R. Dick. 1992. Peptide-inducedconformational changes in class I heavy chains alter major histocompatibilitycomplex recognition.J. Exp. Med. 176:1757.

42. Carbone, F. R., and M. J. Bevan. 1989. Induction of ovalbumin-specific cytotoxicT cells by in vivo peptide immunization.J. Exp. Med. 169:603.

43. Fremont, D. H., M. Matsumura, E. A. Stura, P. A. Peterson, and I. A. Wilson.1992. Crystal structures of two viral peptides in complex with murine MHC classI H-2Kb. Science 257:919.

44. Held, W., J. Roland, and D. H. Raulet. 1995. Allelic exclusion of Ly49 familygenes encoding class I MHC specific receptors on NK cells.Nature 376:355.

45. Moretta, A., M. Vitale, C. Bottino, A. M. Orengo, L. Morelli, R. Argugliaro,M. Barbaresi, E. Ciccone, and L. Moretta. 1993. P58 molecules as putative re-ceptors for major histocompatibility complex (MHC) class I molecules in humannatural killer (NK) cells. Anti-p58 antibodies reconstitute lysis of MHC classI-protected cells in NK clones displaying different specificities.J. Exp. Med.178:597.

46. Ljunggren, H.-G., C. Ohlen, Hoglund, P.,, T. Yamasaki, G. Klein, and K. Karre.1988. Afferent and efferent cellular interactions in natural resistance directedagainst MHC class I deficient tumor grafts.J. Immunol. 140:671.

47. Smith, J. D., W. R. Lie, J. Gorka, N. B. Myers, and T. H. Hansen. 1992. Extensivepeptide ligand exchange by surface class I major histocompatibility complex

molecules independent of exogenous beta 2-microglobulin.Proc. Natl. Acad. Sci.USA 89:7767.

48. Luscher, M. A., B. L. Newton, and B. H. Barber. 1994. Characteristics of het-erologousb2m exchange into H-2Db at the cell surface.J. Immunol. 153:5068.

49. Mandelboim, O., H. T. Reyburn, M. Vales-Gomez, L. Pazmany, M. Colonna,B. Borsellino, and J. L. Strominger. 1996. Protection from lysis by natural killercells of group 1 and 2 specificity is mediated by residue 80 in human histocom-patibility leukocyte antigen C alleles and also occurs with empty major histo-compatibility complex molecules.J. Exp. Med. 184:913.

50. Colonna, M., and J. Samaridis. 1995. Cloning of immunoglobulin-superfamilymembers associated with HLA-C and HLA-B recognition by human natural killercells.Science 268:405.

51. Wagtmann, N., R. Biassoni, C. Cantoni, S. Verdiani, M. S. Malnati, M. Vitale,C. Bottino, L. Moretta, A. Morreta, and E. O. Long. 1995. Molecular clones ofthe p58 NK cell receptor reveal immunoglobulin related molecules with diversityin both and extra- and intracellular domains.Immunity 2:439.

52. Correa, I., L. Corral, and D. H. Raulet. 1994. Multiple natural killer cell-activat-ing signals are inhibited by major histocompatibility complex class I expressionin target cells.Eur. J. Immunol. 24:1323.

53. Ciccone, E., D. Pende, L. Nanni, C. Di Donato, O. Viale, A. Beretta, M. Vitale,S. Sivori, A. Moretta, and L. Moretta. 1995. General role of HLA class I mole-cules in the protection of target cells from lysis by natural killer cells: evidencethat the free heavy chains of class I molecules are not sufficient to mediate theprotective effect.Int. Immunol. 7:393.

766 CLASS I BINDING PEPTIDES AND NK RECOGNITION

by guest on Decem

ber 20, 2015http://w

ww

.jimm

unol.org/D

ownloaded from