Developmental Regulation of a MucinlikeGlycoprotein Selectively Expressed on NaturalKiller CellsBy Eric Vivier,*# J . Michael Sorrell,1 Melissa Ackerly,*Michael J. Robertson,*§ Robert A. Rasmussen,*# Herbert Levine,*and Paul Anderson* II

From the 'Division of Tumor Immunology, Dana-Farber Cancer Institute, and the Departmentsof #Pathology, §Medicine, and 11Rheumatology and Immunology, Harvard Medical School,Boston, Massachusetts 02115, and (Biology Department, Case Western Reserve University,Cleveland, Ohio

Summary

Natural killer (NK) cells are CD3TCR - , CD16+, CD56+ large granular lymphocytes capableofrecognizing and eliminating a variety ofvirus-infected, malignant, and antibody-coated targetcells. Two functionally distinct populations of peripheral blood NK cells can be differentiatedby their surface expression of an isoform of the neural cell adhesion molecule (CD56) . CD56brightNK cells have the attributes of an undifferentiated cell, in that they proliferate in response toexogenous cytokines, but exert poor cytolytic activity. CD56dim NK cells have the attributesof a more differentiated cell, in that they proliferate poorly in response to exogenous cytokines,but are potent rytolytic effector cells. Here we describe the molecular characterization of a NKcell restricted epitope (PEN5) that is selectively expressed on the functionally differentiatedCD56dim NK cells . PEN5+ NK cells proliferate poorly in response to interleukin 2 (IL-2), butare potent cytolytic effectors, whereas PEN5 - NK cells proliferate in response to IL-2, but arepoor cytolytic effectors . Biochemical and immunochemical analyses reveal the PEN5 epitope tobe an unusual sulfated poly-N-lactosamine carbohydrate related to keratan sulfate glycosaminoglycans.Immunoprecipitates prepared using a monoclonal antibody reactive with PEN5 include twopolydisperse membrane-bound glycoproteins, PEN5c i (120-170 kD) and PEN50 (210-245 kD) .Enzymatic deglycosylation reduces the apparent molecular weight of both PEN5 isoforms by80-90%, and classifies PEN50 as a mucinlike glycoprotein . The surface expression of the PEN5epitope is downmodulated by stimuli that induce NK cell proliferation, and it is absent fromleukemic NK cells of patients with granular lymphocyte proliferative disorder. Taken together,these results indicate that PEN5 is a developmentally regulated poly-N-lactosamine epitope associatedwith a mucin-type glycoprotein, whose expression is restricted to the population of nonproliferativeNK cells fully committed to cytolytic effector function .

NK cells are large granular lymphocytes (LGLs)t com-prising 2-15% of PBMC in healthy individuals (1) . Al-

though NK cells do not rearrange or express either of theknown antigen receptor complexes, they can recognize andkill a specific repertoire of virus-infected and transformed cellsin a non-MHC-restricted fashion (2) . With the exception

t Abbreviations used in this paper. ADCC, antibody-dependent cell cyto-toxicity; APC, allophycocyanine; BC, bovine cornea keratan sulfate; BNC,bovine nasal cartilage aggrecan ; CD1, embryonic chick cartilage aggrecan;G1cNac, N-acetyl glucosamine ; GLPD, granular lymphocyte proliferativedisorder; LCM, leucocyte-conditioned medium ; LGL, large granularlymphocyte; N-CAM, neural cell adhesion molecule; PNgase, peptide-Nglycosidase.

2023

of CD16, an Fc receptor for Ig that recognizes Ab-coatedtarget cells (3), the NK cell surface receptors responsible fortarget cell recognition have not been identified . The lack ofa defining surface receptor requires NK cells to be identifiedby a combination ofphenotypic and functional characteristics.Although most peripheral blood NK cells are CD3TCR - ,CD16+, CD56+ (neural cell adhesion molecule [N-CAM])LGL, there is substantial phenotypic and functional hetero-geneity within this population (1) . For example, the surfacedensity of CD56 has been shown to define functionally dis-tinct NK cell populations . CD56bright NK cells are largelyCD16- , agranular lymphocytes deficient in cytolytic effectorfunction, which proliferate vigorously in response to IL-2and resemble embryonic/fetal NK cells (4-6) . By contrast,

J . Exp . Med . C The Rockefeller University Press " 0022-1007/93/12/2023/11 $2.00Volume 178 December 1993 2023-2033

CD56di- NK cells are CD16+ ILL possessing potent cyto-lytic effector function, which proliferate very poorly in re-sponse to IL-2 (4, 5) . Because CD16 and CD56 are not re-stricted to the lymphocyte population (1), and a T cell subsetalso expresses both CD16 and CD56 (1), these moleculescannot define, by themselves, the NK cell population .

To identify cell surface structures more selectively expressedon NK cells, we generated a panel of mouse mAb reactivewith freshly isolated human NK cells . In this article, we re-port the molecular characterization ofPEN5, a sulfated poly-N-lactosamine epitope whose expression is restricted, withinhematopoietic cells, to the functionally differentiated popu-lation of LGL previously characterized as CD3TCR - ,CD56di-, CD16+ cytolytic effectors. Because PEN5 is notexpressed on activated cytotoxic T cells, it can be used todirectly identify this important NK cell subset . We showthat PEN5 + NK cells are potent cytolytic effectors thatproliferate poorly in response to IL-2, whereas PEN5 - NKcells are less potent cytolytic effectors that proliferate stronglyin response to IL-2 . The developmental regulation of PEN5expression suggests that its presence may be required for someaspect of NK cell function .

Materials and MethodsReagents.

Peptide-N-glycosidase (PNgase F) and Endo-cr-N-acetylgalactosaminidase (O-glycanase) were used in the bufferprovided by the manufacturer (Oxford Glycosystems) . KeratanaseI (keratan sulfate 1,4 /3-n-galactanohydrolase ; ICN Biomedicals,Cleveland, OH), keratanase II which attacks oversulfated forms ofkeratan sulfate resistant to keratanase I (Seikagaku America, Inc.,Rockville, MD) and neuraminidase (Calbiochem-NovabiochemCorp., La Jolla, CA) were used in either PBS, PNgase F or O-gly-canase buffers . Chondroitinase ABC (ICN Biomedicals) was usedin sodium acetate 0.05 M, pH 7.4 . Bovine cornea keratan sulfate(BC), as well as other glycoaminoglycans and carbohydrates, werepurchased from Sigma. Trypsin, chymotrypsin, pronase Eand strep-tavidin were also obtained from Sigma Chemical Co. (St . Louis,MO). Allophycocyanine (APC) was obtained from MolecularProbes, Inc. (Eugene, OR).

Antibodies . Mouse mAb reactive with CD2 (SFC13Pt2H9,IgGl), CD3 (SFCIRW2-8C8, IgGI), CD56 (N901, IgGl), andCD20 (H299, IgG2a), as well as isotype-matched control mousemAb (IgG and IgM) were obtained from Coulter Electronics (Hi-aleah, FL). The anti-PEN5 mAb (5H10, IgM) was produced byimmunizing BALB/c mice with digitonin-permeabilized periph-eral blood NK cells (7). Radioiodination of PEN5 mAb was per-formed using Iodobeads (Pierce, Rockford, IL) as previously de-scribed (8) . The characterization of the anti-CD16 mAb (3G8,IgGl), and the anti-keratan sulfate mAb 5134 (IgM) was reportedelsewhere (9, 10) . The following mAb recognize distinct epitopeson most keratin sulfate chains : 1134 (IgG), 2133 (IgG), 3132 (IgM),4131 (IgM), and 8C2 (IgM) (11) . FITC-labeled goat anti-mouseIg (G+M) was purchased from Tago Inc. (Burlingame, CA).

Cells.

All cells were cultured in final medium consisting ofRPMI 1640 (Whittaker Bioproducts, Walkersville, MD) sup-plemented with 10% FCS, 1 mM sodium pyruvate, 2 rnM gluta-mine, and 50 ug/ml gentamycin, all obtained from GIBCO BRL(Gaithersburg, MD). Peripheral blood samples were obtained fromhealthy volunteers . Purified NK and T cells were isolated fromPBMC by negative selection using immunomagnetic bead deple-

2024

PEN5, a NK Cell Lineage-restricted Polylactosamine Epitope

tion (12) . In some experiments, NK cells and cell subsets(CD56h°gh` and CD56d'm) were further purified by cell sorting onan Epics Vflow cytometer (Coulter Electronics) after staining withanti-CD56 mAb. Activation ofNK cells was as described previously(13) . PBMC from three patients with a CD3TCR- , CD16+,CD56+ granular lymphocyte proliferative disorder (GLPD) (14),were isolated by Ficoll-Hypaque gradient centrifugation . The NKcell line NKL was derived from one of these patients (Robertson,M. J., unpublished results) . Splenic B cells were isolated and acti-vated as described (15) . The isolation of peripheral B cells, mono-cytes, and thymocytes was performed using standard methods(16-18) . Peripheral blood platelets were kindly obtained from Dr.E. Weinman (Beth Israel Hospital, NewYork). RBC were obtainedfrom peripheral blood, and granulocytes were purified by Ficoll-Hypaque gradient centrifugation andRBClysis. T4C1 and T4T8C1CTL clones, as well as YTN17, 3.3 NK cell lines, have been de-scribed elsewhere (19-22) . Other allogenic CTL clones were gener-ated by limiting dilutions in rIL2-containing final medium, after8 d of MLR between PBL (5 x 106/ml) and the EBVtransformedlymphoblastoid cell line JY (106/ml) .

Immunoprecipitations.

Cells were resuspended in PBS and sub-jected to radioiodination using "'I by the lactoperoxidase method(23) . Sepharose-bound immune complexes were washed four timesin lysis buffer, and eluted either directly into sample buffer (2%SDS, 10% glycerol, 0.1 MTris-HCI, pH 6.8,0.02% bromophenolblue) before electrophoretic separation, or in elution buffer (0 .15MNH4OH, pH 10.5) before deglycosylation experiments .

Deglycosylation o, fRadioiodinated PEN5 Glycoproteins.

Radioio-dinated PEN5 samples eluted from 5H10-coated Sepharose beadswere dried under vacuum and resuspended in appropriatedeglycosylated enzyme buffers . The following enzymes were usedalone or in combination : PNgase F (310 U/ml), O-glycanase (0 .06U/ml), keratanase I (0 .25 U/ml), and neuraminidase (0 .2 U/ml).Deglycosylation was for 24 h at 37 °C.ELISAforAggrecan-type Proteoglyans.

Wells ofmicrotiter plateswere incubated with 10 ug/ml solutions ofthe indicated aggrecan-type proteoglycans overnight at 4°C. After washing, wells wereincubated with 0.1 MTris, pH 7.6, containing 1% BSA, or withthe indicated enzymes in this buffer. After enzymatic digestion,a standard ELISA was performed using 1:1,000 dilution of 5H10(anti-PEN5) and 5134 (anti-keratan sulfate) mAb, and 1:500 dilu-tion of anti-mouse Ig(G+M) conjugated with alkaline phospha-tase. Color was developed usingp-nitrophenyl phosphate substratein 0.86 M diethanolamine, pH 9.8 . All absorbance values are themean of4 wells (SD <10%) and have been corrected for nonspecificbinding of the second Ab .NK Cell Cytotoxicity and Proliferation Assays.

NKcells were pre-pared from PBMC by negative selection using immunomagneticbeads (12) . These cells were double labeled with rhodamine-conjugated 5H10 (anti-PEN5) and FITC-conjugated anti-CD3 andsorted on an Epics V flow cytometer (Coulter Electronics) to iso-late CD3- PEN5+ and CD3 -PEN5 - subsets . Post-sort analysisconfirmed that the selected markers were >98% present or absentfrom the indicated population . These cells were then assayed fortarget cell killing and IL2-mediated proliferation using standardmethods (5) . The reported stimulation index was calculated as : [av-erage cpm (IL-2 cultures)/average cpm (media cultures)] . An iden-tical analysis was performed using NK cells labeled first with anti-CD56 (NKH1, IgGl) followed by FITC-labeled goat anti-mouseIgG, and second with rhodamine-conjugated 5H10 (anti-PEN5) .These cells were sorted for CD56+PEN5+ and CD56+PEN5 -populations, and assayed for cytotoxicity and proliferation as de-scribed above.

ResultsExpression ofPENS on Hematopoetic Cells.

To identify novelcell surface structures selectively expressed on NK cells, wegenerated a panel of mouse mAb (anti-PEN mAb) that rec-ognized NK cells but not T cells. The hematopoietic expres-sion of one of these mAb (5H10, anti-PEN5) was found tobe restricted to NK cells. As shown in Fig . 1, two-color flowcytometric analysis of PBL revealed PEN5 to be expressedon the majority of CD56+ (Fig. 1, upper left) and CD16+(Fig . 1, upper right) PBL . In contrast, PEN5 was not expressedon CD3+ T cells (Fig. 1, lower left) or on CD20+ B cells(Fig . 1, lower right) . Neither T cell activation induced by mito-genic concentrations of PHA nor Con A, nor B cell activa-tion induced byStaphylococcus aureus Cowan strain I, for 1-6 d,induced the cell surface expression of the PEN5 epitope (Table1) . Similarly, a panel of allogeneic cytotoxic T cell clones(CD3+CD4+CD56 - , CD3+CD8+CD56 - as well asCD3+CD4+CD56+) did not express the PEN5 epitope(Table 2) . Cell surface staining of monocytes, granulocytes,platelets, and erythrocytes also failed to reveal the PEN5 epi-tope (Table 1), confirming that PEN5 is an NK cell-restrictedmolecule. To more precisely analyze the expression of PEN5on NK cells, flow cytometric analysis of PEN5 expressionwas performed on freshly isolated peripheral blood NK cellspurified by negative selection using immunomagnetic beaddepletion (see Materials and Methods) . PEN5 was brightlyexpressed on 71.7 ± 3.5% (mean ± SEM, n = 16) oftheseNK cell preparations whose average phenotype was 75.6 ±

Figure 1 .

PEN5 expression on PBL . PBL were stained by two-colorflow cytometry using the indicated mAbs. Numbers in each quadrant in-dicate the percentage of positively stained cells.

2025

Vivier et al .

Table 1 .

Cell Surface Expression ofPENS on Hematopoietic Cells

Cell type

Relative expression

Peripheral blood T cellsActivated T cells$Thymocytes

Peripheral blood NK cells

+ +NK cell lines : YT.N17

-3.3

±NKL

-

Peripheal blood B cellsSplenic B cellsActivated B cellss

MonocytesGranulocytesPlateletsRBC

" The cell surface expression of 5H10 was assessed by indirect immu-nofluorescence and flow cytometry; (-) <5% positive stained cells ; (±)between 5 and 35% positive-stained cells ; (+ +) >60% positive stainedcells .t Peripheral blood T cells were activated with optimal mitogenic con-centrations of PHA and Con A, and immunofluorescence staining wasperformed at days 2, 4, and 6 after activation .S Splenic B cells were activated with optimal mitogenic concentrationsof Staphylococcus aureus Cowan strain 1, and immunofluorescence stain-ing was performed at days 2, 4, and 6 after activation .

3 .3% CD56+, 74.2 ± 4.0% CD16+, and 8.3 ± 3.5%CD3+ .The phenotypic heterogeneity of peripheral blood NK cells

required a more careful comparison of the relative expressionof PEN5 and CD56. The two-color flow cytometric com-parison shown in Fig . 1, suggested that PEN5 was preferen-tially expressed on the CD56d"" population . This wasconfirmed by comparing the expression of PEN5 on sortedpopulations of CD56d'm and CD56brighl NK cells (Fig . 2) .PEN5 was expressed at a high density on 85.9 ± 2.2% ofCD56dim NK cells (n = 4), and at low density on 31.1 ±5 .3% of CD56bright NK cells . These results indicate thathigh density cell surface expression of the PEN5 epitope isrestricted to the functionally differentiated CD56d'm NKcells . These results also indicate that the cell surface expres-sion of PEN5 defines two distinct subsets of NK cells,PEN5+ and PEN5d'ml - which overlap with the CD56d'mand CD56bright NK cell subsets, respectively. Because the rel-ative expression of CD56 on the surface of NK cells has beenshown to correlate with their proliferative capacity and cyto-lytic effector function (4), we compared the ability of highlypurified populations of PEN5+ and PEN5 - NK cells toproliferate in response to IL-2 and effect natural cytotoxicity.Mononuclear cells enriched for NK cells by magnetic bead

depletion (see Materials and Methods) were sorted intoPEN5+ and PEN5 - cell populations that were either CD3 -(Fig. 3, top) or CD56+ (Fig. 3, bottom) . In both cases,PEN5+ NK cells killed K562 and Molt-4 targets more effec-tively than PEN5 - NK cells. Conversely, the IL-2 inducedproliferation of PEN5 - cells (stimulation index = 14 ± 6[n = 3] in CD3 - sorted cells and 14 ± 8 in CD56+ sortedcells) was significantly greater than that of PEN5+ cells

Unsorted NK cells

CD56 dim NK cells

CD56 bright NK cells

s

Control CD56

Fluorescence Intensity

PEN5

2026

PEN5, a NK Cell Lineage-restricted Polylactosamine Epitope

s

s

PENS

The cell surface phenotype of the indicated T cell clones was performed by immunofluorescence and flow cytometry . (-) <5% positive stainedcells ; (+) >60% positive-stained cells .

(Stimulation index = 2.3 ± 1.7 [n = 3] in CD3 - cells and3.1 ± 1.7 in CD56+ cells) . These results establish thephenotypic and functional similarities between PEN5+ cellsand CD56di- cells, and PEN5' cells and CD56b°eht cells,respectively.PENS Expression Is Downregulated by NK Cell Activa-

tion . CD56dim and CD56b°eht NK cells strongly differ intheir response to proliferative stimuli (4, 5, 24) : CD56dim

23 .

Figure 2 .

Expression of PEN5 on distinct NK cell subsets . Unsorted NK cells, or sorted CD56

and CD56b6sht NK cells were analyzed for theexpression of 5H10 using biotinylated anti-PEN5 mAb and APC-conjugated streptavidin . Controls were performed using mouse isotype-matched con-trol mAb . The numbers in each histogram indicate the percentage of positively stained cells .

Table 2 . Absence of Surface Expression ofPENS on Cytotoxic T Cell Clones

Cell surface expression

Clone CD3 CD2 CD4 C138 CD56

T4Cl "+ +6.5 B4 + +6.5 C1 + +20.1 A2 + +8.17 A + +

20.1 D8 + +T4T8C1 + +

LdaE0

U

Days of culture

0

6

8

10

14

20

K562

MOLT-4

E/T RATIO

Fluorescence Intensity

2027

Vivier et al .

CD56 dim NK cells

CD56 bright NK cells

Control CD56 PEN5 Control CD56 PEN5

1

10100

1

10100

1

10 100

1

16 100

1

16 100

'1

10 160

Figure 3.

Natural cytotoxicity mediated bypurified populations of PEN5+ and PEN5 -NK cells . (Top) NK cells were sorted forCD3-PEN5- (O) and CD3-PEN5' (" )populations. (Bottom) NK cells were sorted forCD56+PEN5- (O) and CD56+PEN5* (" )populations . Results are representative of threeindependent experiments using differentdonors.

Figure 4 .

Kinetics of PEN5 expression onactivated NK cells . Sorted CD56dim andCD56brigh, NK cells were activated for 20 dwith ionomycin and LCM as described (13).At the indicated times, aliquots of the activatedNK cell populations were analyzed for theircell surface phenotype by flow cytometry usingisotype-matched control mAb, anti-CD56 and5H10 mAb. Results indicate the percentage ofpositively stained cells (%) ; the total meanfluorescence intensity is indicated below.

% 75

% 9254

% 99190

z 75

% 9184

% 4949

.

% 33

% 7465

%797101

% 1312

% 7382

% 4340

% 12

% 7971

z s760

z to11

z 6s83

z z8z6

z a4

%8793

I z 86

z 7282

% 1815

% 53

% 8897

% 4028

~ % 11

85112

% 1814

i z 11

% 9 4107

% 1812

z 113

z ez120

z zs11

1~

NK cells proliferate very poorly in response to IL-2, whereasCD56bright NK cells proliferate in response to this stimulus .However, CD56dim NK cells can be induced to proliferatein response to a combination of LCM and ionomycin (24) .We took advantage ofthis observation to correlate PEN5 ex-pression with the proliferative state of sorted CD56dim andCD56bright NK cells . The cell surface expression of CD56and PENS was determined on both NK cell subsets by flowcytometic analysis at the indicated time intervals after activa-tion (Fig. 4) . Activation of CD56d'm NK cells resulted in thetemporal reduction of PENS expression. In parallel, the cellsurface expression of CD56 was temporally increased, andafter 20 d of activation, the cell surface expression of PENSand CD56 on the CD56dim NK cells was largely similar tothat of unactivated CD56b°ght NK cells (i.e., PEN5di-I - andCD56bright). These results are consistent with the absence ofPENS from the cell surface of long-term human NK cell clones(Moretta, A., personal communication) . In addition, PENSwas not expressed on leukemic NK cells (CD3TCR - ,CD16+, CD56+) isolated from patients with GLPD (Fig .5) . Finally, PENS was absent or dimly expressed on threelong-term human NK cell lines, YT.N17, 3.3, NKL (Table1) . These results indicate that PEN5 expression inversely corre-lates with the proliferative capacity of NK cells .

Biochemical Characterization ofthe PEN5 Epitope.

Radio-iodinated lysates prepared from resting NK cells were immu-noprecipitated using the 5H10 (anti-PEN5) mAb or an isotype-matched control mAb . Immunoprecipitates were then sepa-rated under nonreducing conditions on SDS-polyacrylamidegels. As shown in Fig. 6, two diffuse bands were selectivelyimmunoprecipitated by the 5H10 mAb. The average molec-

E

du

Control CD56 PEN5

F-]3

Fluorescence Intensity

2028

Figure 5 .

Cell surface expression of PENS epitope on leukemic NKcells. Peripheral blood NK cells, as well as PBMC isolated from three pa-tients undergoing GLPD blast crisis (GLPD1-3) were analyzed for thecell surface expression of CD56 and PEN5 using indirect immunofluores-cence and flow cytometry. The numbers in each histogram indicate thepercentage of positively stained cells .

PEN5, a NK Cell Lineage-restricted Polylactosamine Epitope

Figure 6 . Immunoprecipita-tion of PENS glycoproteins . De-tergent lysates prepared fromradioiodinated NK cells were im-munoprecipitated using 5H10mAb or mouse IgM control .Samples were then separated undernonreducing conditions on a 6%SDS-polyacrylamide gel .

ular weight of the larger species, PEN5a, was 227 ± 4 kD(n = 12 different donors) . The molecular weight range ofthe polydispersed PEN5a species was 210 ± 3-245 ± 5 kD .The average molecular weight of the smaller species, PEN50was 140 ± 3 kD, with a range of 123 ± 3 to 170 ± 4 W.The migration ofboth PEN5a and a molecules as polydispersebands suggested that they were highly glycosylated. This wasconfirmed in the deglycosylation experiments shown in Fig .7 . PEN5a and /3 were affinity purified from radioiodinatedNK cell lysates using anti-PEN5 mAb and eluted as describedin Materials and Methods. The eluted glycoproteins weretreated with the indicated enzymes before electrophoretic sepa-ration on an SDS-polyacrylamide gel. Compared with themigration of untreated PENS glycoproteins (Fig. 7, lane 1),PNgase F treatment induced the disappearance of PEN5afrom the 210-245-kD range, and the appearance of a deglyco-sylated form of PEN5a migrating at 20-25 kD (Fig. 7, lane6, c2) . In contrast, the apparent mobility of PEN50 was re-duced by only -20 kD after PNgase F incubation . Treat-ment of PEN5 glycoproteins with O-glycanase (Fig . 7, lane3) did not significantly affect their SDS-PAGE migration pat-tern. These results indicate that the PEN5a and PEN50 differmarkedly in their carbohydrate attachment sites, and that80-90% of the apparent molecular weight ofPEN5a is dueto N-linked carbohydrates . Although the recognition of PENSin Western blots was weak, the epitope was expressed onboth PEN5a and PEN50 (data not shown) .The extensive Winked glycosylation ofPEN5a suggested

that it might be a member of one ofthe classes ofglycoproteinscharacterized by such high carbohydrate content (50-90%),such as proteoglycans. Affinity purified radioiodinated PEN5aand (3 molecules were digested by enzymes that cleave the

Figure 7.

Enzymatic deglycosylation ofPENS glycoproteins . Affinity-purified PEN5a and 0 glycoproteins were eluted from antibody-coatedSepharose beads, then subjected to deglycosylation for 24 h at 37°C usingPNgase F (lane 6), O-glycanase (lane 3), keratanase I (lane 2), O-glycanaseand keratanase (lane 4), neuraminidase (lane 5), and PNgase F and neur-aminidase (lane 7) and appropriate buffers. Control eluates incubated inPBS without any enzymes were separated in lane 1 . Samples were sepa-rated under nonreducing conditions on a 6-12% SDS-polyacrylamide gra-dient gel .

glycosaminoglycan side chains from proteoglycan core pro-teins. Chondroitinase ABC, heparitinase and heparinase didnot affect the migration pattern of PEN5a or PEN5a (datanot shown) . By contrast, incubation ofPENS molecules withkeratanase I reduced the apparent molecular weight ofPEN5afrom 210-245 to 35-40 kD (Fig. 7, lane 2, c1) . It is likelythat the difference between the PNgase F-digested (35-40kD) c1 core protein (Fig. 7, lane 2) and the keratanase-digested(25-30 kD) c2 core protein (Fig. 7, lane 4), results from amore complete deglycosylation of the PEN5a glycoprotein .Whereas keratanase treatment only slightly reduced the poly-dispersity of PEN5a, the combination of O-glycanase andkeratanase I treatment reduced the apparent molecular weightof PEN5a from 120-170 to 25-30 kD (Fig. 7, lane 4) . Takentogether, these results indicate that PEN5a contains 80-90%

2029

Vivier et al .

N-linked keratanase I-sensitive carbohydrates, whereas PEN5acontains 80-90% O-linked carbohydrates and keratanase I-sen-sitive chains. This extensive O-linked glycosylation is consis-tent with PEN5a being a mucinlike glycoprotein . In addi-tion, treatment with neuraminidase induced a slight reductionin the polydispersity, as well as a shift in the apparent molec-ular weight ofboth PEN5a and a, indicating that sialic acidresidues are also present on both glycoproteins (Fig . 7, lane5) . Treatment of PENS glycoproteins with a combinationof PNGase F and neuraminidase (Fig . 7, lane 7), resulted inthe same effect that PNGase F had alone, confirming the pres-ence ofterminal sialic acid residues on Winked carbohydratespresent on PEN5a. The cl and c2 deglycosylated forms ofPEN5a and a proteins were not immunoprecipitable by the51110 mAb (data not shown), indicating that the epitope rec-ognized by the anti-PENS mAb contains keratanase I-sensi-tive carbohydrate chains .

Reactivity ofAnti-PENS mAb with Keratan Sufate Glycos-aminoglycans . To test whether the anti-PENS mAb wasdirectly raised against keratan sulfate carbohydrates, we ex-amined the effect ofexogenous keratan sulfate carbohydrateson the binding of PENS mAb to NK cells. Radioiodinated51110 (anti-PENS) mAb was combined with various con-centrations of BC, and the mixture was then incubated withNK cells. As shown in Fig. 8 A, the binding of radiolabeled51110 mAb to NK cells was inhibited by BC keratan sulfatein a dose-dependent manner. Preincubation of NK cells withthe same concentrations of BC keratan sulfate followed bywashes, did not affect the binding of 51110 mAb (data notshown), indicating that the anti-PENS mAb reacted withcarbohydrate determinants present on keratan sulfate glycos-aminoglycans . Incubation of anti-PENS mAb with simplesugars or other glycosaminoglycans was without any effect(see legend to Fig. 8) . Furthermore, treatment of NK cellswith keratanase I induced a 58.5% ± 8.4 (n = 4) decreasein the reactivity of 51110 mAb with NK cells (Fig . 8 B) .Parallel treatment of NK cells with chondroitinase ABC orneuraminidase did not have any effect on 51110 reactivity.Interestingly, the 51110 epitope was totally insensitive totrypsin and chymotrypsin but was removed by pronase E treat-ment . Finally, an ELISA was used to compare the bindingof 51110 (anti-PENS) and 5D4 (anti-keratan sulfate) to thekeratan sulfate/chondroitin sulfate proteoglycans expressed incartilage from different species . As shown in Fig. 8 C (top),the 51110 mAb (solid ban) recognized aggrecan-type pro-teoglycans derived from embryonic chick cartilage (CD1, top)and from bovine nasal cartilage (BNC, middle) . As a positivecontrol, the anti-keratan sulfate mAb 5D4 (Fig . 8, open ban)also reacted with untreated CDl and BNC, whereas its reac-tivity with keratanase-treated samples was reduced. TreatmentofCD1 and BNC with either keratanase I or II, reduced 51110reactivity. Treatment ofCDl and BNC with chondroitinaseABC is known to increase the expression of keratan sulfateepitopes (11) . Consequently, digestion ofCDl and BNC withchondroitinase ABC increased the binding of both 51110 and5D4 . As a negative control, neither mAb recognized theSwarm rat chondrosarcoma aggrecan (RC), which does notcontain keratan sulfate (Fig. 8 C, bottom) . Although 5D4 also

Figure 8 .

Reactivity of anti-PEN5 mAb with keratan sulfate glycosaminoglycans. (A) 1 125-labeled 5H10 mAb (106 cpmAample) was preincubatedfor 20 min at 4°C in PBS in the presence of the indicated concentrations ofBC. The mixture was then added to NK cells for another 20-min incubationat 4°C, before three washes in PBS-1% BSA. Samples were counted in a .y-counter, and results are expressed as mean cpm of duplicate samples (SD<10%). When used in incubation with NK cells or anti-PEN5 mAb, the following carbohydrates used at 10 mg/ml were without any effect on 5H10binding to NK cell surface: chondroitin sulfate B, heparin, heparan sulfate, dextran sulfate, GlcNAc, mannose 6-phosphate, lactose, galactose-6-phosphate,fucose, glucose 6-phosphate, glucose, and galactose. (B) Peripheral blood NK cells were incubated in PBS-1% BSA for 3 h or 45 min at 37°C withglycosidases (0.025 U/ml) or proteases (5 fag/ml), respectively. Cell surface expression of PEN5 epitope was then analyzed by flow cytometry using5H10 mAb. Percent modulation was calculated as the ratio of the total linear mean fluorescence intensity of the treated cells over that of untreatedcontrol cells. (C) The antigenicity of 5H10 mAb for aggrecan proteoglycans was analyzed by ELISA as described in Materials and Methods . The anti-keratan sulfate mAb 5D4 was used as a positive control. Chondroitinase ABC was used at 0.04 U/ml, keratanase I was used at 0.05 U/ml, and keratanaseII was used at 0.004 U/ml, for 1 h at 37 °C .

2030

PEN5, a NK Cell Lineage-restricted Polylactosamine Epitope

reacted with the keratan sulfate proteoglycan isolated fromshark cranial cartilage (SHK), 5H10 did not . These resultsindicate that the PENS epitope is present in some, but notall, keratan sulfate chains, and is therefore distinct from thestandard sulfated poly-N-lactosamine repeat sequence,Gala1-4(sulfated) G1cNAc . This result is consistent with theinability of six distinct anti-keratan sulfate mAbs (1B4, 2D3,3D2, 4D1, 8C2, and 5D4, data not shown) to bind to NKcells. Preliminary results (Vivier, E., and P. Anderson, manu-script in preparation) using immunoelectron microscopy showthat the PENS epitope is located some distance from the plasmamembrane (43.4 + 12.8 nm [n = 50]), suggesting that, likeother cell surface mucins, the membrane-bound glycopro-teins carrying the PENS epitope are extended threadlikeproteins.

DiscussionWe describe the identification and molecular characteriza-

tion ofPENS, a novel cell surface epitope selectively expressedon peripheral blood NK cells. Several lines of evidence indi-cate that the PENS epitope is an unusual sulfated poly-N-lactosamine carbohydrate that can be expressed on keratansulfate glycosaminoglycans. First, keratan sulfate glycosamino-glycans selectively compete with PENS molecules for bindingto the 5H10 (anti-PENS) mAb. Second, treatment of NKcells with keratanase I reduces the cell surface expression ofthe PENS epitope. This enzyme attacks sites where there isan unsulfated Gal unit, which may be adjacent to a 6-sulfatedG1cNAc unit ; completely unsulfated disaccharides cannot serveas substrates for this enzyme (25) . Third, the 5H10 mAbrecognizes epitopes located on keratan sulfate chains fromaggrecan-type proteoglycans obtained from embryonic chickand bovine nasal cartilages. These keratan sulfateglycosaminoglycans contain, as their prominent structure, arepeating Galal-4G1cNac, disaccharides that can be sulfatedon either, or both saccharide units . Failure of the anti-PENSmAb to recognize an epitope on a third aggrecan-type pro-teoglycan, from shark cranial cartilage, suggests that this mAbdoes not recognize a structure common to all keratan sulfatechains . Further modification of keratan sulfate chains can re-sult from differential fucoslyation, sialylation, or perhaps ad-dition of other carbohydrates, such as mannose to the basicrepeating structure (26) . It is possible that the PENS epitopemight contain one, or more, unusual structures, which couldaccount for its presence in some, but not all, keratan sulfatechains.

Immunoprecipitation of NK cell detergent lysates with theanti-PENS mAb, revealed two distinct glycoproteins, PEN5aand PEN5a. Enzymatic deglycosylation indicated. that bothspecies were 80-90% carbohydrate by weight . This resultraises the possibility that these proteins are either proteoglycansor mucinlike glycoproteins . Proteoglycans are high molec-ular weight glycoproteins in which specific glycosaminoglycansare bound to proteins via Gal-xylose-Ser linkages, with theexception of keratan sulfate as discussed below (27) . PENSmolecules are free of xylose-linked carbohydrates, since the

2031

Vivier et al.

cell surface expression of PENS, and the SDS-PAGE migra-tion ofboth PEN5a and a molecules are not affected by treat-ments with chondroitinase ABC (Fig. 8 B), heparitinase,heparinase, or p-nitrophenyl a-n-xylopyranoside (data notshown) . However, the reactivity of anti-PENS mAbs withsulfated poly-N-lactosamine carbohydrates present on keratansulfate glycosaminoglycans raises the possibility that PEN5aand/or PEN5a glycoproteins may be cell surface-associatedkeratan sulfate proteoglycans. Keratan sulfate maybe Winkedor O-linked to core proteoglycan proteins (26) . To date,O-linked keratan sulfate chains have been identified only incartilage, bone, and brain (26) . In the cornea, keratan sulfatechains are Winked to Asn residues on 37- and 25-kD coreproteins (26) . The linkage oligosaccharide is related to bian-tiennary oligosaccharides found in complex-type glycopro-teins, with the keratan sulfate chain extending from one branchand sialic acid terminating the second branch (28) . Antigeni-cally similar core proteins, with similar molecular weight,have been identified in a variety oftissues (29) . However, thereappears to be considerable variability in the glycosylation ofthese keratan sulfatelike molecules. The PEN5a glycopro-tein shares many of the same characteristics as these keratansulfatelike molecules . By contrast, PEN5a is an O-linked gly-coprotein containing keratanase I-sensitive carbohydrates . Ithas been reported that mucin-type glycoproteins secreted bycultured hamster tracheal epithelial cells contain keratanaseI sensitive poly-N-lactosamine carbohydrates (30) . Mucinlikeglycoproteins are highly glycosylated proteins containing amajority ofO-linked oligosaccharides, and are associated withthe cell membrane in a number ofcell types (31) . Classificationof PEN5a as a NK cell-specific membrane-bound mucinlikeglycoprotein is most consistent with our data . Therefore, ifPEN5a and PEN5a are associated, the PEN5a-Pen5a com-plex may be analogous to the ASGP-1-ASGP-2 complex de-rived from ascitic mammary adenocarcinoma cells (32), inwhich only one component (ASGP-1) of the complex is amucin-like glycoprotem. The precise biochemical identificationof the carbohydrates present on PEN5a and PEN5a glyco-proteins, as well as the amino acid sequence of the core pro-teins, will allow definitive classification .Our results clearly show that the cell surface expression

of this unique sulfated poly-N-lactosamine carbohydrate isrestricted, within the hematopoetic compartment, to NK cells.It is intriguing that other cell surface carbohydrates, such asthe acidic polylactosamine carbohydrate determinants CD57(HNK-1) and sialyl-Lewisx-i in humans, and asialo-GM1ganglioside in mice, have previously been identified on thesurface of NK cells (1) . Taken together, these results suggestthat a general characteristic of NK cells is their expressionof several membrane-bound poly-N-lactosamine carbohydrates .The selective expression of the PENS epitope on restingCD56d'°t NK cells, and its downregulation in response toNK cell proliferation, strongly suggest that PENS is a markeroffunctionally differentiated NK cells that are specialized forcytolytic effector functions. It is possible that the absence ofPENS expression on CD56bagbt, NK cells and GLPD cellsreflects a developmentally regulated shift in the activities of

glycosyltransferases involved in keratan sulfate metabolism .Modulation of glycosyltransferase activities induced by celldifferentiation and activation has been described in severalcell types including T lymphocytes (33, 34).

Some of the biochemical features of PENS glycoproteinspoint the way for the study of their function . First, carbohy-drates are major mediators ofcell-cell interactions (35) . Second,in their protease-resistance as well as their extended rodlikestructure, in addition to the high extent of PEN50 O-glyco-sylation, the PENS glycoproteins resemble epithelial cellmucins . Since the mucin-type glycoproteins serve a protec-tive role on the epithelial cell surface, it is tempting to specu-late that PENS glycoproteins protect NK cells from their own

References

cytolytic machinery. The selective expression ofsuch shieldingproteins on the terminally differentiated subset of NK cellswould be consistent with the concomitant acquisition offullycompetent cytotoxic functions. Exogenous mucins have beenshown to inhibit NK cell killing (36), as well as ADCC medi-ated by eosinophils (37) supporting their potential involve-ment in resistance to NK cell cytolytic functions (36) . Finally,glycosaminoglycans expressed on cell surface proteoglycanshave been shown to capture growth factors (i.e., TGF-a andbasic fibroblast growth factor), as well as proadhesive cytokines(i .e ., macrophage inflammatory protein 1/3), and it remainsto be investigated whether, on NK cells, PENS glycopro-teins could also exert such functions (38, 39).

The authors thank Drs. J.-L . Taupin, A. S. Freedman, andN. Rochet for helpful advice; Drs. S. F. Schlossmanfor continual encouragement and support; Drs. H. S. Slayter and Xu for production of the electron micro-graphs ; S. Lazo, J. Daley, and C. Groves for flow cytometric studies .

This work was supported by grants from the National Institutes of Health (CA-53595) . E. Vivier wassupported by a postdoctoral fellowship from the Arthritis Foundation. P. Anderson was supported bya Pew Scholar Award and an Investigator Award from the Cancer Research Institute .

Address correspondence to Dr. Eric Vivier, Centre d'Immunologie, Marseille-Luminy, INSERM-CNRS,Case 906, Marseille 13288, France.

Received for publication 19 April 1993 and in revised form 10 September 1993.

2032 PENS, a NK Cell Lineage-restricted Polylactosamine Epitope

1 . Trinchieri, G. 1989 . Biology of natural killer cells . Adv. Im- 9 . Perussia, B., and G. Trinchieri. 1984 . Antibody 3G8, specificmunol. 47 :187 . for the human neutrophil Fc receptor, reacts with natural killer

2. Lanier, L.L ., andJ.H . Phillips. 1992 . Natural killer cells . Curr. cells . J. Immunol. 132:1410 .Opin . Immunol. 4:38. 10 . Caterson, B., J.E . Christner, andJ.R. Baker. 1983 . Identification

3 . Ravetch, JV, and B. Perussia. 1989 . Alternative membrane of a monoclonal antibody that specifically recognizes cornealforms of Fcy RIII (CD16) on human natural killer cells and and skeletal keratan sulfate. Monoclonal antibodies to carti-neutrophils : cell type-specific expression oftwo genes that differ lage proteoglycan . J. Biol. Chem . 258:8848.in single nucleotide substitutions. J. Exp Med. 170:481 . 11 . Sorrell, J., and M. Caterson . 1990 . Monoclonal antibodies

4. Nagler, A., L.L . Lanier, S. Cwirla, and J.H . Phillips . 1989. specific for keratan sulfate detect epithelial-associated carbohy-Comparative studies ofhuman FcyRIII-positive and negative drates. Histochemistry. 94:269.natural killer cells. J Immunol. 143:3183. 12 . Vivier, E., N. Rochet, M. Ackerly, J. Petrini, H. Levine, J.

5. Baume, D.M., M.J . Robertson, H. Levine, T.J . Manley, P.W. Daley, and P. Anderson . 1992 . Signaling function of recon-Schow, andJ. Ritz . 1991 . Differential responses to interleukin-2 stituted CD16:r:y receptor complex isoforms . Int. Immunol.define functionally distinct subsets ofhuman natural killer cells. 4:1313 .Eur. J. Immunol. 22 :1 . 13 . Robertson, M.J ., T.J . Manley, C. Donahue, H. Levine, and

6. Phillips, J.H., T Hori, A. Nagler, N. Bhat, H. Spits, and L.L . J. Ritz . 1993 . Costimulatory signals are required for optimalLanier. 1992 . Ontogeny of human natural killer (NK) cells: proliferation ofhuman natural killer cells.J. Immunol 150:1705 .fetal NK cells mediate cytolytic function andexpress cytoplasmic 14 . Oshimi, K. 1988 . Granular lymphocyte proliferative disorders :CD3e, S proteins. J. Exp Med. 175 :1055 . report of 12 cases and review of the literature. Leukemia. 2:617 .

7 . Anderson, P., M.L . Blue, C. O'Brien, and S.F. Schlossman. 15 . Freedman, A.S., J.M . Munro, G.E . Rice, M.P. Bevilaqua, C.1989b. Monoclonal antibodies reactive with the TCR z chain: Morimoto, B.W McIntyre, K. Rhynhar, J. Pober, and L.M .Production and characterization using a new method . J Im- Nadler. 1990 . Adhesion of human Bcells to germinal centersmunol. 143:1899. in vitro involves VLA-4 and INCAM-110. Science (Wash. DC).

8. Vivier, E., P. Morin, C. O'Brien, B. Drucker, S.F. Schlossman, 249:1030.and P. Anderson . 1991a. Tyrosine phosphorylation of the Fc 16 . Leprince, C., S. Cohen-Kaminsky, S. Berrih-Aknin, B. Vernet-y RIII (CD16) : t complex in human natural killer cells : in- Der Garabedian, D. Treton, P Galanaud, andY Richard. 1990 .duction by antibody dependent cytotoxicity but not by nat- Thymic B cells from myasthenia gravis patients are activatedural killing . J. Immunol. 146:206 . B cells . J. Immunol. 145:2115 .

17 . Salem, P., S. Deryckx, A. Dulioust, E. Vivier, Y. Denizot,C. Damais, C.A . Dinarello, and Y. Thomas. 1990 . Im-munoregulatory functions of paf-acether. IV. Paf-acether en-hances interleukin-1 (IIA) production by muramyl dipeptide-stimulated monocytes. J. Immunol. 144:1338.

18 . Vivier, E., P. Morin, Q. Tian, M.-L . Blue, S.F. Schlossman,and P. Anderson . 1991 . Expression and tyrosine phosphoryla-tion of the T cell receptor ~-subunit in human thymocytes .J. Immunol. 146:1142 .

19 . Anderson, P., M. Blue, and S. Schlossman . 1988c. Comodula-tion of CD3 and CD4: evidence for a specific association be-tween CD4 and N5% of the CD3T cell receptor complexeson helper T lymphocytes. J. Immunol. 140:1732 .

20 . Blue, M.L ., D.A . Hafler, J.F. Daley, H. Levine, K.A . Craig,J.B. Breitmeyer, and S.F. Schlossman . 1988 . Regulation of Tcell clone function via CD4 and CD8 molecules .J. Immunol.140:376 .

21 . Yodoi, J., K. Teshigawara, T Nikaido, K. Fukui, T Noma,T. Honjo, M. Takigawa, M. Sasaki, N. Minato, M. Tsudo,et al. 1985 . TCGF (IL2)-receptor inducing factor(s) . I . Regulation of IL-2 receptor on a natural killer-like cell line (YTcells) . J. Immunol. 134:1623 .

22 . Kornbluth, J., N. Flomenberg, and B. Dupont . 1982 . Cell sur-face phenotype of a cloned line of human natural killer cells.J. Immunol. 129:2831.

23 . Vivier, E., N. Rochet, J.-P. Kochan, D.H. Presky, S.F.Schlossman, andP. Anderson . 1991 . Structural similarities be-tween Fc receptors and T cell receptors: expression of the .ysubunit of FceRI in human T cells, NK cells, and thymocytes .J. Immunol . 147:4263.

24 . Robertson, M.J ., R.J . Soiffer, S.F. Wolf, T.J . Manley, C.Donahue, D. Young, S.H . Herrmann, andJ. Ritz . 1992 . Re-sponse of natural killer (NK) cells to NK cell stimulatory factor(NKSF) : cytolytic activity and proliferation of NK cells isdifferentially regulated by NKSF. J. Exp. Med. 175 :779.

25 . Ito, M., Y. Hirabashi, and T. Yamagata. 1986 . Substratespecificity of endo-b-galactosidases from Flavobacterium ker-atolyticus and Esterichia freundii is different from that ofPseudo-monas sp . J. Biochem . 100:773 .

2033

Vivier et al.

26 . Funderburgh, J., M. Funderburgh, M. Mann, and G. Conrad .1991 . Physical and biological properties of keratan sulfate pro-teoglycans . Biochem . Soc. Trans. 19 :871 .

27 . Hook, M., L. Kjellen, S. Johansson, and J. Robinson . 1984 .Cell-surface glycosaminoglycans . Annu . Rev. Biochem. 53:847 .

28 . Oeben, M., R. Keller, H. Stuhlsatz, and H. Greiling. 1987 .Constant and variable domains of different disacharide struc-tures in corneal sufate chains . Biochem . J. 248:85 .

29 . Funderburgh, J.L ., B. Caterson, andG.W. Conrad . 1987 . Dis-tributio n ofproteoglycans antigenically related to corneal ker-atan sulfate proteoglycan . J. Biol. Chem . 262:11634 .

30 . Wu, C., C. Plopper, and P. Cheng. 1991 . Mucin-like glyco-protein secreted by cultured hamster tracheal epithelial cells .Biochem . J. 277:713 .

31 . Strous, G.J ., and J. Dekker. 1992 . Mucin-type glycoproteins.Crit. Rev. Biochem. Mol. Biol. 27:57.

32 . Sherblom, A.P., andK.L . Carraway. 1980. A complex of twocell surface glycoproteins from ascites mammary adenocarci-noma cells . J. Biol. Chem . 255:12051 .

33 . Kuiijpers, T.W. 1993 . Terminal glycosyltransferase activity : aselective role in cell adhesion . Blood. 81 :873 .

34 . Piller, F., V Piller, R.I . Fox, and M. Fukuda . 1988 . HumanTlymphocyte activation is associated with changes in O-glycanbiosynthesis. J. Biol. Chem . 263:15146 .

35 . Jessel, TM., M.A. Hynes, andJ. Dodd . 1990 . Carbohydratesand carbohydrate-binding proteins in the nervous system. Annu .Rev. Neurosci. 13 :227 .

36 . Ogata, S., P.J . Maimonis, and S.H . Itzkowitz. 1992 . Mucinsbearing the cancer-associated sialosyl-Tn antigen mediate in-hibition of natural killer cytotoxicity. Cancer Res . 52:4741.

37 . Hayes, D.F., D.S. Silberstein, S.W. Rodrique, and D.W. Kufe .1990. DF3 antigen, a human epithelial cell mucin, inhibits adhe-sion of eosinophils to antibody-coated targets. J. Immunol.145 :962.

38 . Ruoslahti, E., andY Yamaguchi. 1991 . Proteoglycans as modu-lators of growth factor activities. Cell. 64:867 .

39 . Tanaka, Y, D.H . Adams, S. Hubscher, H. Hirano, U. Sieben-list, and S. Shaw. 1993 . T-cell adhesion induced by proteoglycan-immobilized cytokine MIP-1(3. Nature (Land.). 361:79.