of April 4, 2018.This information is current as

Migratory CapacitySubsets with Distinct Responsiveness and CD27 Dissects Mature NK Cells into Two

Yoshihiro Hayakawa and Mark J. Smyth

http://www.jimmunol.org/content/176/3/1517doi: 10.4049/jimmunol.176.3.1517

2006; 176:1517-1524; ;J Immunol 

Referenceshttp://www.jimmunol.org/content/176/3/1517.full#ref-list-1

, 14 of which you can access for free at: cites 43 articlesThis article

        average*  

4 weeks from acceptance to publicationFast Publication! •    

Every submission reviewed by practicing scientistsNo Triage! •    

from submission to initial decisionRapid Reviews! 30 days* •    

Submit online. ?The JIWhy

Subscriptionhttp://jimmunol.org/subscription

is online at: The Journal of ImmunologyInformation about subscribing to

Permissionshttp://www.aai.org/About/Publications/JI/copyright.htmlSubmit copyright permission requests at:

Email Alertshttp://jimmunol.org/alertsReceive 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 © 2006 by The American Association of1451 Rockville Pike, Suite 650, Rockville, MD 20852The American Association of Immunologists, Inc.,

is published twice each month byThe Journal of Immunology

by guest on April 4, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

by guest on April 4, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

CD27 Dissects Mature NK Cells into Two Subsets with DistinctResponsiveness and Migratory Capacity1

Yoshihiro Hayakawa2 and Mark J. Smyth

Lineage differentiation and the formation of heterogeneous mature subsets are crucial for immune cells to maintain a breadth ofresponsiveness to pathogens while controlling reactivity to self. In this study, we report that CD27 is a key marker of the NK celllineage, dissecting the mature Mac-1high NK cell pool into two functionally distinct subsets. The CD27low NK cell subset possessesa higher threshold to stimulation and appears to be tightly regulated by the expression of NK cell inhibitory receptors. Compar-atively, the CD27high NK cell subset displays a greater effector function, exhibits a distinct tissue distribution and responsivenessto chemokines, and interacts productively with dendritic cells. Importantly, we have verified that CD27high and CD27low subsetswith distinct cell surface phenotypes also exist in human peripheral blood. These findings clearly reclassify mature NK cells intotwo distinct subsets and begin to discern their specific role in immune responses. The Journal of Immunology, 2006, 176: 1517–1524.

A lthough NK cells lack the ability to generate Ag-specificreceptors by gene rearrangement, NK cells distinguishabnormal cells by using a repertoire of Ig-like and C-

type lectin receptors that deliver a finely tuned balance of inhibi-tory and activating signals (1–3). A number of these receptorsrecognize ligands on transformed and virus-infected cells and thusstimulate NK cell cytokine production (e.g., IFN-�) (4), secretionof cytotoxic granules (5), and expression of TNF superfamilydeath-inducing ligands (6) that effectively control virus infectionand tumor initiation and spread (7–9). Furthermore, NK cells canalso regulate subsequent components of the adaptive immune sys-tem in host protection from tumors or viruses, via their capacity tocross talk with APCs like dendritic cells (DCs)3 (10–17).

The multiple stages of NK cell development have been proposedon the basis of their function, phenotype, and proliferative capac-ities (18–21). In the mouse, the earliest lineage-committed precur-sors are characterized by expression of the IL-2R and IL-15R com-mon � subunit (IL-2/IL-15R� or CD122). At the next stage ofmaturation, there is sequential acquisition of NK1.1 and CD94-NKG2 receptors and the integrin �v subunit. NK cells then expressLy-49 and c-kit, followed by an NK cell expansion stage that ischaracterized by an up-regulation of DX5 and down-regulation ofthe integrin �v subunit. During the perceived final stage of NK cellmaturation, NK cells up-regulate Mac-1 (CD11�) and CD43 andacquire effector function, producing high levels of IFN-� and ex-

erting high levels of cytotoxicity (22). We have recently demon-strated that TRAIL identifies immature mouse NK cells duringadult life, and it is the dominant cytotoxic effector molecule ex-pressed by NK cells in fetal mice (23). These observations gainedfrom in vivo analysis suggest that mouse NK cell development ismuch more complicated and sophisticated than the simple picturethat in vitro NK cell cultures might predict.

The subdivision of NK cells into functional subsets was origi-nally proposed in the early 1980s by Lanier et al. (24), and thenrevisited in late 1990s with the discovery that CD56 expressionlevels distinguished functionally distinct human NK cell subsets(25–27). It has been shown that human CD56bright and CD56dim

NK cells are distinct in their functions, including cytotoxicity, cy-tokine production, and migratory capacity (27). However, CD56 isnot expressed in rodents, and there has been no similar clear evi-dence for functionally distinct mature NK cell subsets in mice.Thus, until now it has been very difficult to relate our knowledgeof mouse NK cell biology with human NK cell biology and trans-late this information into clinical practice.

In this study, we show in mice that the mature Mac-1high NKcell pool can be further dissected into two functionally distinctCD27high and CD27low subsets. In concert with distinct patterns ofsurface receptor expressed by these two NK cell subsets, theirprototypic NK cell effector functions, proliferative capacity, tissueorganization, interactions with DCs, and response to chemokinescan be clearly distinguished. Furthermore, CD27 can dissect hu-man peripheral blood NK cells into similar two subsets that pro-vide a new definition distinct from CD56. Thus, this study pro-vides the first functional dissection of mature NK cell subsets inthe mouse and begins to provide a platform from which bothmouse and human NK cells can be explored in immune responseand disease.

Materials and MethodsMice

Inbred wild-type C57BL/6 (B6) mice were purchased from The Walter andEliza Hall Institute of Medical Research. CD45.1 congenic C57BL/6 micewere purchased from the Animal Resources Centre. B6 RAG-1�/� micewere bred and maintained at the Peter MacCallum Cancer Centre. Allexperiments were performed according to animal experimental ethics com-mittee guidelines.

Cancer Immunology Program, Trescowthick Laboratories, Peter MacCallum CancerCentre, East Melbourne, Victoria, Australia

Received for publication September 15, 2005. Accepted for publication November2, 2005.

The costs of publication of this article were defrayed in part by the payment of pagecharges. This article must therefore be hereby marked advertisement in accordancewith 18 U.S.C. Section 1734 solely to indicate this fact.1 This work was supported by a Cancer Research Institute postdoctoral fellowship (toY.H.) and National Health and Medical Research Council of Australia research fel-lowships and a program grant (to M.J.S.).2 Address correspondence and reprint requests to Dr. Yoshihiro Hayakawa, CancerImmunology Program, Trescowthick Laboratories, Peter MacCallum Cancer Centre,St. Andrews Place, East Melbourne, Victoria 3002, Australia. E-mail address:yoshihiro.hayakawa@petermac.org3 Abbreviations used in this paper: DC, dendritic cell; IP-10, IFN-�-inducible protein-10; I-TAC, IFN-inducible T cell � chemoattractant; KLRG1, killer cell lectin-likereceptor G1; LN, lymph node.

The Journal of Immunology

Copyright © 2006 by The American Association of Immunologists, Inc. 0022-1767/06/$02.00

by guest on April 4, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

Reagents

Abs to TCR� (H57-597), NK1.1 (PK136), pan NK cells (DX5), CD122(TM-�1), CD11b (M1/70), CD27 (LG.3A10), CD69 (H1.2F3), CD45R/B220 (RA3-6B2), CD94 (18d3), Ly-49C/I (5E6), Ly-49D (4E5), Ly-49G2(4D11), Ly-49I (YLI-90), NKG2D (CX5), NKG2A/C/E (20d5), NKG2AB6

(16a11), killer cell lectin-like receptor G1 (KLRG1; 2F1), CD43 (S7),CD127 (A7R34), CD62L (MEL-14), c-kit (2B8), CXCR3 (220803), CCR5(C34-3448), CXCR4 (2B11/CXCR4), and CD45.1 (A20) were purchasedform BD Pharmingen, R&D Systems, and eBioscience. Purified anti-mouse CCR7 was kindly provided by J. Swirner (Gottingen, Germany).

Flow cytometry

Mononuclear cells from the spleen, liver, bone marrow (BM), lymph node(LN), blood, and lung were isolated, as described. For staining NK cells,mononuclear cells were first preincubated with CD16/32 (2.4G2) mAb toavoid the nonspecific binding of Abs to Fc�R. Then the cells were incu-bated with a saturating amount of mAbs. Flow cytometric analysis wasperformed with an LSR II instrument (BD Biosciences).

BM chimeras

BM cells were isolated from B6 CD45.1 congenic mice and transferred(2 � 106) into sublethally irradiated (5 Gy) wild-type B6 mice (CD45.2).Some mice were injected with 500 �g of anti-CD70 (FR70) at days �1, 0,1, and 3, and then every 3 days. Cells from spleen were harvested at theindicated time points after transfer and stained for CD45.1, NK1.1, TCR�,Mac-1, and CD27. NK cells from donor BM origin were determined byelectronic gating on NK1.1� TCR�� CD45.1� cells.

In vivo BrdU uptake assay

Mice were injected i.p. with 2 mg of BrdU (Sigma-Aldrich) twice per day(10–12 h apart). On the day of analysis, cells were harvested from BrdU-injected mice, and at least two non-BrdU-treated control mice were in-cluded with each experiment as negative control. Cells were stained forNK1.1, TCR�, Mac-1, and CD27, and then subsequently stained for in-tracellular incorporation of BrdU (BrdU Flow Kit; BD Biosciences), ac-cording to manufacturer’s instruction. Some mice were injected daily with500 �g of anti-CD70 (FR70) together with BrdU injection. This dose hasbeen shown to neutralize CD70 function in vivo (12).

Cytotoxicity assay

Cytotoxic activity was assessed against Yac-1 or RMA-Rae-1� target cellsby a standard 51Cr release assay. Effector cells were isolated from RAG-1�/� spleen and purified Mac-1highCD27high and Mac-1highCD27low NKcells by cell sorting (�95% purity). Target cells were labeled with 100�Ci/ml Na2

51CrO4 for 60 min at 37°C, and labeled target cells (104/well)were incubated in a total volume of 200 �l with effector cells in 96-wellU-bottom plates. The plates were centrifuged before incubation, and after4 h the supernatant was harvested and counted in a gamma counter.

Cytokine assay

For in vitro NK cell culture, NK cells were isolated from B6 RAG-1�/�

spleen and purified Mac-1highCD27high and Mac-1highCD27low NK cells bycell sorting (�95% purity). Cells (5 � 104/well) were stimulated withIL-12 (100 ng/ml; R&D Systems) and/or IL-18 (100 ng/ml; R&D Sys-tems), or BM-derived DCs (1:1 ratio) in RPMI 1640 medium with humanrIL-2 (500 U/ml; Chiron). BM-derived DCs were generated with culturesupernatants from GM-CSF- and IL-4-producing cell lines, as previouslydescribed (28). After 24-h incubation, the cell-free supernatants were har-vested and subjected to ELISA. The amounts of IFN-� (BD Pharmingen)were quantitated by specific sandwich ELISAs.

Chemotaxis assay

Chemotactic migration activity was performed, as previously described(29). Briefly, RAG-1�/� splenocytes (106) or purified NK cells (105) wereplaced in the upper chamber of Transwell inserts (8-�m pore size; CorningCostar). For assays with purified NK cells, spleen cells were isolated fromB6 RAG-1�/� and purified Mac-1highCD27high and Mac-1highCD27low NKcells by cell sorting (�95% purity), and then placed in the upper chamber.Inserts were placed in wells containing medium alone (control) or mediumcontaining recombinant chemokine. The chemokines secondary lymphoidtissue chemokine/CCL21, IFN-�-inducible protein-10 (IP-10)/CXCL10,IFN-inducible T cell � chemoattractant (I-TAC)/CXCL11, MIP-1�/CCL3,and stromal cell-derived factor 1/CXCL12 were kindly provided by S.McColl (University of Adelaide, Adelaide, South Australia, Australia). Af-ter incubation and harvest, the number of cells migrating to the bottomchamber were counted. With B6 RAG-1�/� splenocytes, the cells werefurther subjected to staining with Abs and quantified by flow cytometry.

Statistical analysis

Data were analyzed for statistical significance using the Mann-Whitney Urank sum test. Values of p �0.05 were considered significant.

ResultsCD27 dissects mature NK cells into two major subsets

Generally, lymphocyte differentiation into specific subsets and/orlong-lived, Ag-specific effector/memory cells is critical for effec-tive adaptive immunity. In particular, interactions between a vari-ety of TNFR superfamily and TNF superfamily members havebeen shown to play a key role in promoting lymphocyte survival,and the expression level of TNF receptor super family correlateswith the maturity of T cells and B cells (30). Interestingly, pro-portion of naive NK cells constitutively expresses CD27, and wehave shown previously that stimulation via CD27 triggers NK cellactivation, cytokine production, and induction of subsequent adap-tive immune responses (12, 31). We initially investigated the ex-pression of CD27 on naive NK cells in the context of known NKcell maturation and differentiation markers (Fig. 1). Surprisingly,CD27 expression dissected the mature spleen Mac-1high NK cellpopulation into Mac-1highCD27high and Mac-1highCD27low NKcells. CD27low NK cells expressed a mature CD122� DX5high Ly-49s� CD43highMac-1high NK cell phenotype, clearly demonstrat-ing dissection of the mature NK cell pool by CD27 expression.

Surface phenotype of NK cell subsets

Because it had been previously suggested that NKR expressionwas tightly linked to the maturation of NK cells (3, 32, 33), wenext examined the expression of NKR and other maturation andactivation molecules on immature Mac-1low and mature Mac-1high

NK cell subsets. In this respect, there were clear differencesamong the NK cell subsets defined as Mac-1lowCD27high,Mac-1highCD27high, and Mac-1highCD27low expression (Fig. 2).The Mac-1highCD27low NK cell subset demonstrated higher pro-portions of cells expressing Ly-49C and I (which recognize selfMHC class I molecules in C57BL/6 mice) and KLRG1. By con-trast, a reduced proportion of CD94/NKG2high cells was observed

FIGURE 1. CD27 dissects the mature Mac-1high NK cell pool into two subsets. Cells isolated from B6 mice spleen were stained for the NK cellmaturation markers (CD122, DX5, Ly-49s: Ly-49A/C/D/G2/I, CD43, Mac-1), as indicated, together with NK1.1, TCR�, and CD27. The dot plots shownare profiles for a NK maturation marker and CD27 expression on electronically gated NK1.1�TCR�� cells. Data are representative of at least threeexperiments.

1518 NK CELL SUBSETS WITH DISTINCT FUNCTIONS

by guest on April 4, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

in the Mac-1highCD27low NK cell subset (Fig. 2). Furthermore, theactivation/differentiation marker expression was also distinctamong these NK cell subsets, with Mac-1high NK cell subsets dis-playing higher CD62L expression, and a proportion of CD27high

NK cells constitutively expressing CD69. Interestingly,Mac-1lowCD27high NK cells specifically expressed IL-7R�(CD127) and higher level of c-kit, confirming that Mac-1lowCD27high

NK cells may represent an earlier stage of NK cell development.

Distinct tissue distribution of mature NK cell subsets

The tissue distribution of the lymphocyte subsets has been con-sidered diverse and may be important for tissue-resident/specificimmune responses in general. Therefore, the tissue distribution ofNK cell subsets in vivo may provide clues as to their involvementand role in immune responses in their microenvironment. Thus, weexamined the distribution of Mac-1/CD27 NK cell subsets in var-ious lymphoid organs and other tissues, including BM, spleen,LN, peripheral blood, liver, and lung (Fig. 3). Surprisingly,Mac-1highCD27low NK cells were comparatively excluded fromLN, whereas this subset was the dominant tissue-resident NK cellpopulation within lung (Fig. 3). Interestingly, Mac-1highCD27low

NK cells were also the dominant population circulating within pe-ripheral blood (Fig. 3). In agreement with previous observations,

BM NK cells mainly consisted of a mixture of immatureMac-1lowCD27high NK cells and Mac-1highCD27high NK cells. Im-portantly, these NK cell subsets faithfully maintained their distinctcell surface phenotype (as shown for spleen in Fig. 2) in all of theorgans we have examined (data not shown), suggesting that thisdefinition of NK cell subsets is not microenvironment dependent.

Delayed appearance of mature Mac-1highCD27low NK cells

We next examined the kinetics of the appearance of mature NKcell subsets in a BM chimera assay using B6 Ly-5.2 congenic mice(Fig. 4). Upon congenic BM transfer into sublethally irradiated B6Ly-5.1 mice, donor-derived Ly-5.2� NK cells developed from thetransferred BM progenitor cells and expressed an immature Mac-1lowCD27high NK cell phenotype 1 wk posttransfer (Fig. 4a). At 2wk post-BM, the appearance of Mac-1highCD27high NK cells wasobserved, and then Mac-1highCD27low NK cells finally began toappear by 4 wk post-BM transfer. Eight weeks after BM transfer,a dominant proportion of Mac-1highCD27low NK cells wasobserved in the spleens of host mice, indicating that theMac-1highCD27low NK cell subset appears as a relatively late stageof NK cell differentiation after BM reconstitution. Furthermore,continuous neutralization of CD70 by Ab did not affect NK cellsubset formation following BM transfer (Fig. 4b), suggesting that

FIGURE 2. Characterization ofsurface phenotype of NK cell sub-sets. Cells isolated from the spleensof B6 mice were stained for the cellsurface markers, as indicated, to-gether with NK1.1, TCR�, Mac-1,and CD27. The histogram plotsshown are expression levels for theindicated cell surface markers on NKcell subsets (R1:Mac-1low CD27high,R2:Mac-1highCD27high, R3:Mac-1highCD27low) electronically gated onthe relevant population, as shown.Data are representative of at leastthree experiments.

1519The Journal of Immunology

by guest on April 4, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

the CD27-CD70 interaction may not be critical for NK cell subsetformation in vivo.

Mature CD27high NK cells predominantly proliferate in vivo

A recent study demonstrated that a large proportion of BM NKcells was in cell cycle and was most likely the precursors of thenondividing or slowly dividing spleen NK cells (34, 35). We haveconducted an analysis that indicates that mature Mac-1highCD27high

NK cells are the predominant population in both the spleen (Fig.5a) and BM (data not shown) that actively take up BrdU togetherwith immature Mac-1lowCD27high NK cells (data not shown) (22).Thus, the CD27low subset has a very limited capacity for prolif-eration and turnover among NK cell subsets (Fig. 5a). Consis-tently, Mac-1highCD27low NK cells were the predominant propor-tion of NK cells that survived upon administration of hydroxyurea,a cell cycle-specific cytotoxic agent (data not shown), implyingthat Mac-1highCD27low NK cells may represent a long-lived/se-nescent subset of NK cells. Furthermore, continuous inhibition ofCD70 did not affect this NK cell turnover, as determined by in vivouptake of BrdU (Fig. 5b), indicating that the CD27-CD70 interac-tion is not involved in maintaining the homeostasis of NK cellsin vivo.

Functionally distinct NK cell subsets

Because self MHC-recognizing inhibitory Ly-49 receptors arethought to prevent autoaggression by NK cells, a Mac-1highCD27low

NK cell subset that expresses a more abundant level of self MHC-recognizing Ly-49s (Ly-49 C and I) might be expected to reactdistinctly with self vs nonself cells. Indeed, after purifying the Mac-1highCD27high and Mac-1highCD27low subsets by cell sorting, theMac-1highCD27high NK cell subset demonstrated its distinct cytotoxiccapacity against MHC class I-, NKG2D ligand-expressing RMA-Rae-1� target tumor cells (Fig. 5c). By contrast, the CD27low subsetdisplayed only background levels of cytotoxicity against RMA-Rae-1� cells. The cytotoxic capacity of Mac-1highCD27low NK cellsagainst the class I MHC-mismatched NK cell-sensitive (NKG2Dligand-expressing) Yac-1 target cells was relatively lower (2-fold)than that mediated by Mac-1highCD27high NK cells (Fig. 5d), whileNKG2D expression was similar between the populations (Fig. 2).Interestingly, Mac-1highCD27high NK cells displayed non-NKG2D-dependent cytotoxicity against Yac-1 target cells, suggesting thatMac-1highCD27high NK cells can kill target cells by an activationpathway distinct from NKG2D (Fig. 5d).

Early cytokine production by NK cells in response to a varietyof APC (e.g., DC)-derived factors, such as IL-12 and IL-18, isknown to be critical for effective early NK cell control of infectionand the induction of subsequent adaptive immune responses.Therefore, we next examined the potential of mature NK cell sub-sets to produce cytokine in response to IL-12 and IL-18 in vitro.Surprisingly, only Mac-1highCD27high NK cells produced detect-able amounts of IFN-� in response to either IL-12 or IL-18 alone(Fig. 5e). Although Mac-1highCD27low NK cells produced IFN-�

FIGURE 3. Distinct tissue distribution of NKcell subsets. Cells isolated from the bone marrow(a), spleen (b), lymph node (c), liver (d), periph-eral blood (e), and lung (f) of B6 mice werestained for NK1.1, TCR� Mac-1, and CD27. Thedot plots shown are profiles for Mac-1 andCD27 expression on electronically gatedNK1.1� TCR�� cells. The numbers represent thepercentage of cells � SD in each quadrant. Dataare representative of at least five experiments.

1520 NK CELL SUBSETS WITH DISTINCT FUNCTIONS

by guest on April 4, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

upon stimulation with both IL-12 and IL-18, considerably higheramounts of IFN-� were produced by Mac-1highCD27high NK cells(Fig. 5e). These results clearly indicated that Mac-1highCD27high

NK cells possessed a greater ability to produce IFN-�, comparedwith Mac-1highCD27low NK cells, in response to cytokine stimu-lation. Recently, human and mouse NK cells have been shown toengage in a productive cross talk with DCs, resulting in NK cellIFN-� secretion (15, 16). Importantly, Mac-1high CD27high NKcells also displayed a far greater ability to produce IFN-�, com-pared with CD27low NK cells, when they were cocultured withDCs (Fig. 5e), strongly suggesting that Mac-1highCD27high NKcells may play a predominant role in NK:DC cross talk in lym-phoid organs, where they are the predominant NK cell population(Fig. 3).

Distinct chemokine sensitivity of NK cell subsets

A recent study has demonstrated the importance of chemokines forNK cell recruitment to LN, and this NK cell recruitment appears tobe involved in subsequent Th1 immune responses through NK cellIFN-� production (17). In this context, we comprehensively ex-amined the expression of chemokine receptors on NK cell subsetsas determined by Mac-1 and CD27 expression. Notably, we founddistinct chemokine receptor expression on mature NK cell subsets(Fig. 6a). Importantly, Mac-1high CD27high NK cells constitutivelyexpress CXCR3, whereas there was no detectable CXCR3 on Mac-1highCD27low NK cells among mature NK cell pool. Both of themature NK cell subsets demonstrated a very broad level of expres-sion of CXCR4, while neither CCR5 nor CCR7 expression wasdetected (Fig. 6a). Furthermore, we have also confirmed functionalevidence for a distinct chemokine sensitivity of mature NK cellsubsets using an in vitro chemotaxis assay. In response to CXCR3ligands (IP-10, I-TAC), Mac-1highCD27high NK cells showed spe-cific chemotactic activity, whereas Mac-1highCD27low NK cellsdid not (Fig. 6b). Both of the mature NK cell subsets responded toCXCR4 ligand (stromal cell-derived factor 1), although there wereslight differences in their responsiveness (Fig. 6b). Thus, NK cellsubsets may be functionally divided based on their distinct migra-tory capacity and their chemokine receptor expression. Specificchemotaxis of the NK cell subsets was confirmed by assays usingIP-10 or secondary lymphoid tissue chemokine and purified Mac-1highCD27high and Mac-1highCD27low NK cells (Fig. 6c).

DiscussionNK cells are well recognized as one of the first lines of host de-fense against pathogens and tumors (1). Recent evidence stronglysuggested that NK cells are not only effector cells, but also im-portant in regulating the adaptive immune response through theirinteractions with APCs, particularly DCs (14, 15). Despite a recentexpansion in our knowledge defining unique developmental stagesof the NK cell lineage (3, 18), mature NK cell differentiation isvery poorly understood.

The concept of dissecting mature NK cells into functionally dis-tinct subpopulations has previously been pursued in humans. Hu-man NK cells can be subclassified into two functionally distinctsubsets, CD56bright and CD56dim NK cells (20, 27). HumanCD56bright NK cells are considered to be poorly cytotoxic, butpotent cytokine producers, whereas CD56dim NK cells expresshigher levels of cytotoxicity, but are poor cytokine producers. Ithas also been shown that CD56bright NK cells express distinct lev-els of chemokine receptors and adhesion molecules (26) and mayaccumulate within inflammatory sites (36). Because a functionalhomologue of the human CD56 molecule does not exist in themouse, it has been impossible to discern the relationships betweenthese human NK cell subsets and mature NK cells studied in themouse. Consequently, the mouse studies have been limited in theirimpact on the clinical monitoring of human NK cell function andthe design of improved therapeutics for human disease.

Our analysis has now clearly dissected the mature mouse NKcell pool into at least two major subsets, Mac-1highCD27high andMac-1highCD27low NK cells. Most importantly, each of these NKcell populations displayed a distinct NKR expression profile thatcorrelated with their functions. In particular, it was striking thatMac-1highCD27low NK cells displayed skewed expression of in-hibitory Ly-49 receptors (C and I isoforms) that recognize selfMHC class I molecule on the B6 background (H-2b). Compara-tively, the Mac-1highCD27low NK cell population contained agreater proportion of Ly-49 C- and I-expressing NK cells, whereasLy-49G2-expressing NK cells were similarly distributed in bothMac-1highCD27high and Mac-1highCD27low NK cell subsets. Sim-ilar distinctions in killer Ig-related receptor expression were notedon CD27high and CD27low human NK cell subsets (data notshown). There was also a significant difference in the constitutive

FIGURE 4. Delayed appearanceof Mac-1high CD27low after BM re-constitution. a, BM cells were iso-lated from B6 CD45.1 congenic miceand transferred into sublethally irra-diated (5 Gy) wild-type B6 mice(CD45.2). b, Some mice were in-jected with 500 �g of anti-CD70(FR70) at days �1, 0, 1, and 3, andthen every 3 days. Cells from spleenwere harvested at the indicated timepoints after transfer and stained forCD45.1, NK1.1, TCR�, Mac-1, andCD27. NK cells from donor BM or-igin were determined by electronicgating on NK1.1�TCR��CD45.1�

cells. The numbers represent the per-centage of cells � SD in each quad-rant. Data are representative of twoexperiments.

1521The Journal of Immunology

by guest on April 4, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

expression of the KLRG1 between the mouse NK cell subsets,with only Mac-1highCD27low NK cells expressing abundant levelsof KLRG1. Interestingly, it has been reported that following virus-induced proliferation/expansion of NK cells and virus-specificCD8� T cells in vivo, the responsiveness of effector cells is im-paired and such senescent effector cells expressed higher levels ofKLRG1 (37, 38). We have demonstrated that the KLRG1-express-ing, Mac-1highCD27low NK cells can be distinguished from the NKcells incorporating BrdU in vivo. Consistently, Mac-1high CD27low

NK cells were resistant to a cell cycle-sensitive cytotoxic agent;therefore, they may represent long-lived/senescent NK cells.

Functionally, Mac-1highCD27high and Mac-1highCD27low NKcells displayed distinct cytotoxicity and cytokine production. Mac-1highCD27high NK cells showed much greater responsiveness toactivatory ligand expressed on tumor cells (NKG2D and non-NKG2D pathway), and demonstrated effective cytotoxicity againstsuch tumor target cells even in the presence of MHC class I ex-pression. Moreover, Mac-1highCD27high NK cells responded toIL-12 and IL-18, key cytokines derived from professional APCs,by rapidly producing IFN-�. Conversely, Mac-1highCD27low NKcells displayed very low or no responsiveness to the same cellular(NKG2D ligand) and cytokine (IL-12/IL-18) stimulation. Surpris-ingly, Mac-1highCD27low NK cells were not able to override in-hibitory signals from their Ly-49 receptors even when they recog-nized ligands to NKG2D. Mac-1highCD27low NK cells could be

stimulated to make IFN-� with a combination of IL-12 and IL-18;however, the amount of IFN-� produced was still far lower thanthat of Mac-1highCD27high NK cells, suggesting that in a compet-itive environment the Mac-1highCD27low cells would be relativelynonresponsive. In addition, Mac-1highCD27high NK cells produceddetectable amounts of GM-CSF upon IL-12 and IL-18 stimulation,whereas GM-CSF production was not detectable from Mac-1highCD27low cells (data not shown). Neither TNF-�, IL-10, norIL-13 was detectable upon IL-12/IL-18 stimulation of mature NKcell subsets, although such stimulation may not be optimal to in-duce these cytokines from NK cells (data not shown). It now re-mains to examine the production of other NK cell-derived cyto-kines under the appropriate stimulatory conditions. Furthermore,Mac-1highCD27high NK cells showed greater responsiveness toDCs. Mac-1highCD27high NK cells produced higher amount ofIFN-� upon in vitro coculture with BM-derived DCs, comparedwith Mac-1highCD27low cells. A similar observation has been re-ported for the human CD56bright NK cell subset (39). Consideringthe preferential distribution of Mac-1highCD27high NK cells inlymphoid organs, this particular NK cell subset may have a pre-dominant role in cross talk with other immune cells, particularlyDCs. It remains unclear which type of cells might interact withCD27low NK cells, but their expression of MHC class I andNKG2D ligands will be an important determinant of the outcome.

FIGURE 5. Different homeostaticproliferation and functions of matureNK cell subsets. a, B6 mice were in-jected i.p. with 2 mg of BrdU twiceper day (10–12 h apart) for the indi-cated time period. Cells were har-vested from BrdU-injected B6 miceand stained for NK1.1, TCR�, andCD27, and then subsequently stainedfor intracellular incorporation ofBrdU. b, Some mice were injecteddaily with 500 �g of anti-CD70(FR70) together with BrdU injection.This dose has been shown to neutral-ize CD70 function in vivo (12). Cellswere harvested from BrdU-injected B6mice and stained for NK1.1, TCR�,Mac-1, and CD27, and then subse-quently stained for intracellular incor-poration of BrdU. Data are representa-tive of two experiments. c and d, Mac-1highCD27high and Mac-1highCD27low

NK cells were sorted from B6 RAG-1�/� spleen cells. Cytotoxicity wasdetermined against RMA-Rae-1� (c)or Yac-1 (d) or target cells in stan-dard 4-h 51Cr release assay. To in-hibit NKG2D function, anti-NKG2D (C7) (30 �g/ml) was addedto the assay. Data are representativeof two experiments. e, Sorted NK cellsubsets (5 � 104 cells/well) werestimulated with IL-12 (100 ng/ml),IL-18 (100 ng/ml), IL-12/IL-18 (100ng/ml), or BM-derived DCs (1:1 ra-tio) in the presence of IL-2 (500U/ml), and cell-free culture superna-tants were subjected to IFN-�ELISA. Data are representative oftwo experiments.

1522 NK CELL SUBSETS WITH DISTINCT FUNCTIONS

by guest on April 4, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

It also remains to be determined whether distinct monokine re-sponsiveness was simply due to their distinct receptor expressionor to distinct responsiveness, such as intracellular signaling path-way, of each subset to those monokines.

Collectively, the specific functional characteristics and tissuedistribution displayed prompts us to propose a novel classificationof mature NK cell subpopulations. We contend there are two ma-ture Mac-1high NK cell subsets: 1) CD27high NK cells that have alower activation threshold and are predominantly involved inNK:DC cross-talk, and 2) CD27low NK cells that have a greaterrestriction by self MHC, are long-lived, and preferentially resideamong nonlymphoid tissues.

The functional distinctions between the CD27high and CD27low

NK cell subsets also correlated with very distinct chemokine re-sponsiveness and in vivo distribution of these NK cell subsets.Surprisingly, we have now shown that Mac-1highCD27low NK cellsare normally excluded from the LN and lack CXCR3 expression(Fig. 6) (17). By contrast, CD27high NK cells were the predomi-nant proportion of NK cells resident among the LN NK cells andexpressed high constitutive levels of CXCR3. CD27high NK cellsdisplayed active chemotaxis toward CXCR3 ligands (CXCL10;IP–10, CXCL11; I-TAC) that are generally induced by IFN-� (40),implying that the CXCR3-CXCR3 ligand pathway may be a feed-back regulatory loop for NK cell IFN-� production, and furtherrecruit this effector type NK cells into sites of IFN-� production.Interestingly, we demonstrated that Mac-1highCD27low NK cellswere the predominant NK cell subset in lung and peripheral blood.Considering Mac-1highCD27low NK cells are the predominant pro-portion among NK cell subsets in peripheral blood, the NK cell

distribution in lung tissue may largely depend on vascular supply.Therefore, if the Mac-1highCD27low NK cell subset is distributed inall nonlymphoid tissues, it is possible this subset plays a distinctsurveillance/patrolling role outside of the lymphoid tissueenvironment.

It has been recognized that there is clear, but lineage distinct,correlation of CD27 expression with lymphocyte differentiation(30, 41, 42). Furthermore, it has been demonstrated recently thatthe interaction between CD27 and CD70 plays an important role inhemopoiesis (43). Although our data suggest that the CD27-CD70interaction does not appear to be critical for the appearance ofmature NK cells from BM precursors, it will be crucial to deter-mine the molecular factors that control formation and differentia-tion of mature NK cell subset in vivo.

Importantly, we have verified that CD27high and CD27low sub-sets with distinct cell surface phenotypes also exist in human pe-ripheral blood (data not shown), and therefore, CD27 is also anovel mature subset marker in the human NK cell lineage. Inagreement with observations in the mouse, we detected in a num-ber of healthy donors two distinct CD27high and CD27low sub-populations among human peripheral blood CD3�CD161� NKcells. Evidently, like human CD56bright NK cells, mouse CD27high

NK cells are effective producers of IFN-�. In the mouse, we havedemonstrated that both monokines IL-12 and IL-18, and DCsthemselves, induce more IFN-� from the CD27high NK cell subset.What is more perplexing is that, in contrast to mouse CD27high NKcells, human CD56bright NK cells are described as a poorly cyto-toxic subset. Further studies will now be required to determine allthe functional distinctions between human CD27high and

FIGURE 6. Chemokine receptor expression and sensitivity of NK cell subsets. a, Cells isolated from B6 mice spleen were stained for the indicatedchemokine receptors together with NK1.1, TCR�, Mac-1, and CD27. The histogram plots shown are expression levels for chemokine receptors on NK cellsubsets (Mac-1highCD27high, Mac-1highCD27low) electronically gated on relevant population, as shown. The numbers represent the percentage of cells withpositive staining. Data are representative of at least three experiments. b, Cells (106) harvested from B6 RAG-1�/� spleen were added to upperchamber, and the indicated chemokines (200 ng/ml) were added to the lower chamber. The cultures were incubated for 3 h to assess chemotacticability, and cells migrating into the lower chamber were harvested, counted, and stained with NK1.1, TCR�, Mac-1, and CD27. c, Mac-1highCD27high

and Mac-1highCD27low NK cells were sorted from B6 RAG-1�/� spleen cells. Cells (105) were added to upper chamber, and the indicated chemokines(200 ng/ml) were added into the lower chamber. The cultures were incubated for 3 h to assess chemotactic ability; the cells migrating into the lowerchamber were harvested; and chemotaxis was determined by cell count. Data are representative of two experiments.

1523The Journal of Immunology

by guest on April 4, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

CD56bright NK cell subsets. Nevertheless, by discovering CD27 asa common marker for human and mouse NK cell subsets, our studyprovides a very important platform on which to study mature NKcells in immune responses in experimental animals. It is very likelythis may provide a rapid conduit of information from the mousethat can be potentially applied in the study of human NK cellsfollowing vaccination and in a variety of human infectious dis-eases, autoimmunity, and cancer.

AcknowledgmentsWe gratefully thank Stephen Nutt for his very helpful discussions, andShaun McColl for providing reagents. We thank Mark Shannon,Rachel Cameron, Shannon Griffiths, Ralph Rossi, Andrew Fryga, and thestaff of the Peter MacCallum Cancer Centre for their generous support.

DisclosuresThe authors have no financial conflict of interest.

References1. Lanier, L. L. 2005. NK cell recognition. Annu. Rev. Immunol. 23: 225–274.2. Karre, K. 1997. How to recognize a foreign submarine. Immunol. Rev. 155: 5–9.3. Raulet, D. H., R. E. Vance, and C. W. McMahon. 2001. Regulation of the natural

killer cell receptor repertoire. Annu. Rev. Immunol. 19: 291–330.4. Biron, C. A., and L. Brossay. 2001. NK cells and NKT cells in innate defense

against viral infections. Curr. Opin. Immunol. 13: 458–464.5. Trapani, J. A., and M. J. Smyth. 2002. Functional significance of the perforin/

granzyme cell death pathway. Nat. Rev. Immunol. 2: 735–747.6. Smyth, M. J., K. Takeda, Y. Hayakawa, J. J. Peschon, M. R. van den Brink, and

H. Yagita. 2003. Nature’s TRAIL: on a path to cancer immunotherapy. Immunity18: 1–6.

7. Yokoyama, W. M., and B. F. Plougastel. 2003. Immune functions encoded by thenatural killer gene complex. Nat. Rev. Immunol. 3: 304–316.

8. Smyth, M. J., Y. Hayakawa, K. Takeda, and H. Yagita. 2002. New aspects ofnatural-killer-cell surveillance and therapy of cancer. Nat. Rev. Cancer 2:850–861.

9. Cerwenka, A., and L. L. Lanier. 2001. Natural killer cells, viruses and cancer.Nat. Rev. Immunol. 1: 41–49.

10. Fernandez, N. C., A. Lozier, C. Flament, P. Ricciardi-Castagnoli, D. Bellet,M. Suter, M. Perricaudet, T. Tursz, E. Maraskovsky, and L. Zitvogel. 1999.Dendritic cells directly trigger NK cell functions: cross-talk relevant in innateanti-tumor immune responses in vivo. Nat. Med. 5: 405–411.

11. Gerosa, F., B. Baldani-Guerra, C. Nisii, V. Marchesini, G. Carra, andG. Trinchieri. 2002. Reciprocal activating interaction between natural killer cellsand dendritic cells. J. Exp. Med. 195: 327–333.

12. Kelly, J. M., P. K. Darcy, J. L. Markby, D. I. Godfrey, K. Takeda, H. Yagita, andM. J. Smyth. 2002. Induction of tumor-specific T cell memory by NK cell-me-diated tumor rejection. Nat. Immunol. 3: 83–90.

13. Mocikat, R., H. Braumuller, A. Gumy, O. Egeter, H. Ziegler, U. Reusch,A. Bubeck, J. Louis, R. Mailhammer, G. Riethmuller, et al. 2003. Natural killercells activated by MHC class Ilow targets prime dendritic cells to induce protec-tive CD8 T cell responses. Immunity 19: 561–569.

14. Raulet, D. H. 2004. Interplay of natural killer cells and their receptors with theadaptive immune response. Nat. Immunol. 5: 996–1002.

15. Degli-Esposti, M. A., and M. J. Smyth. 2005. Close encounters of different kinds:dendritic cells and NK cells take center stage. Nat. Rev. Immunol. 5: 112–124.

16. Moretta, L., G. Ferlazzo, M. C. Mingari, G. Melioli, and A. Moretta. 2003.Human natural killer cell function and their interactions with dendritic cells.Vaccine 21(Suppl. 2): S38–S42.

17. Martin-Fontecha, A., L. L. Thomsen, S. Brett, C. Gerard, M. Lipp,A. Lanzavecchia, and F. Sallusto. 2004. Induced recruitment of NK cells tolymph nodes provides IFN-� for T(H)1 priming. Nat. Immunol. 5: 1260–1265.

18. Yokoyama, W. M., S. Kim, and A. R. French. 2004. The dynamic life of naturalkiller cells. Annu. Rev. Immunol. 22: 405–429.

19. Raulet, D. H. 1999. Development and tolerance of natural killer cells. Curr. Opin.Immunol. 11: 129–134.

20. Colucci, F., M. A. Caligiuri, and J. P. Di Santo. 2003. What does it take to makea natural killer? Nat. Rev. Immunol. 3: 413–425.

21. Perussia, B., Y. Chen, and M. J. Loza. 2005. Peripheral NK cell phenotypes:multiple changing of faces of an adapting, developing cell. Mol. Immunol. 42:385–395.

22. Kim, S., K. Iizuka, H. S. Kang, A. Dokun, A. R. French, S. Greco, andW. M. Yokoyama. 2002. In vivo developmental stages in murine natural killercell maturation. Nat. Immunol. 3: 523–528.

23. Takeda, K., E. Cretney, Y. Hayakawa, T. Ota, H. Akiba, K. Ogasawara,H. Yagita, K. Kinoshita, K. Okumura, and M. J. Smyth. 2005. TRAIL identifiesimmature natural killer cells in newborn mice and adult mouse liver. Blood 105:2082–2089.

24. Lanier, L. L., A. M. Le, J. H. Phillips, N. L. Warner, and G. F. Babcock. 1983.Subpopulations of human natural killer cells defined by expression of the Leu-7(HNK-1) and Leu-11 (NK-15) antigens. J. Immunol. 131: 1789–1796.

25. Sedlmayr, P., L. Schallhammer, A. Hammer, M. Wilders-Truschnig,R. Wintersteiger, and G. Dohr. 1996. Differential phenotypic properties of humanperipheral blood CD56dim� and CD56bright� natural killer cell subpopulations.Int. Arch. Allergy Immunol. 110: 308–313.

26. Frey, M., N. B. Packianathan, T. A. Fehniger, M. E. Ross, W. C. Wang,C. C. Stewart, M. A. Caligiuri, and S. S. Evans. 1998. Differential expression andfunction of L-selectin on CD56bright and CD56dim natural killer cell subsets.J. Immunol. 161: 400–408.

27. Cooper, M. A., T. A. Fehniger, and M. A. Caligiuri. 2001. The biology of humannatural killer-cell subsets. Trends Immunol. 22: 633–640.

28. Hayakawa, Y., V. Screpanti, H. Yagita, A. Grandien, H. G. Ljunggren,M. J. Smyth, and B. J. Chambers. 2004. NK cell TRAIL eliminates immaturedendritic cells in vivo and limits dendritic cell vaccination efficacy. J. Immunol.172: 123–129.

29. Ebert, L. M., and S. R. McColl. 2002. Up-regulation of CCR5 and CCR6 ondistinct subpopulations of antigen-activated CD4� T lymphocytes. J. Immunol.168: 65–72.

30. Watts, T. H. 2005. TNF/TNFR family members in costimulation of T cell re-sponses. Annu. Rev. Immunol. 23: 23–68.

31. Takeda, K., H. Oshima, Y. Hayakawa, H. Akiba, M. Atsuta, T. Kobata,K. Kobayashi, M. Ito, H. Yagita, and K. Okumura. 2000. CD27-mediated acti-vation of murine NK cells. J. Immunol. 164: 1741–1745.

32. Dorfman, J. R., and D. H. Raulet. 1998. Acquisition of Ly49 receptor expressionby developing natural killer cells. J. Exp. Med. 187: 609–618.

33. Roth, C., J. R. Carlyle, H. Takizawa, and D. H. Raulet. 2000. Clonal acquisitionof inhibitory Ly49 receptors on developing NK cells is successively restricted andregulated by stromal class I MHC. Immunity 13: 143–153.

34. Prlic, M., B. R. Blazar, M. A. Farrar, and S. C. Jameson. 2003. In vivo survivaland homeostatic proliferation of natural killer cells. J. Exp. Med. 197: 967–976.

35. Jamieson, A. M., P. Isnard, J. R. Dorfman, M. C. Coles, and D. H. Raulet. 2004.Turnover and proliferation of NK cells in steady state and lymphopenic condi-tions. J. Immunol. 172: 864–870.

36. Dalbeth, N., R. Gundle, R. J. Davies, Y. C. Lee, A. J. McMichael, andM. F. Callan. 2004. CD56bright NK cells are enriched at inflammatory sites andcan engage with monocytes in a reciprocal program of activation. J. Immunol.173: 6418–6426.

37. Voehringer, D., C. Blaser, P. Brawand, D. H. Raulet, T. Hanke, and H. Pircher.2001. Viral infections induce abundant numbers of senescent CD8 T cells. J. Im-munol. 167: 4838–4843.

38. Robbins, S. H., K. B. Nguyen, N. Takahashi, T. Mikayama, C. A. Biron, andL. Brossay. 2002. Cutting edge: inhibitory functions of the killer cell lectin-likereceptor G1 molecule during the activation of mouse NK cells. J. Immunol. 168:2585–2589.

39. Vitale, M., M. Della Chiesa, S. Carlomagno, C. Romagnani, A. Thiel, L. Moretta,and A. Moretta. 2004. The small subset of CD56brightCD16� natural killer cellsis selectively responsible for both cell proliferation and interferon-� productionupon interaction with dendritic cells. Eur. J. Immunol. 34: 1715–1722.

40. Farber, J. M. 1997. Mig and IP-10: CXC chemokines that target lymphocytes.J. Leukocyte Biol. 61: 246–257.

41. Van Baarle, D., S. Kostense, M. H. van Oers, D. Hamann, and F. Miedema. 2002.Failing immune control as a result of impaired CD8� T-cell maturation: CD27might provide a clue. Trends Immunol. 23: 586–591.

42. McHeyzer-Williams, L. J., and M. G. McHeyzer-Williams. 2005. Antigen-spe-cific memory B cell development. Annu. Rev. Immunol. 23: 487–513.

43. Nolte, M. A., R. Arens, R. van Os, M. van Oosterwijk, B. Hooibrink,R. A. van Lier, and M. H. van Oers. 2005. Immune activation modulates hema-topoiesis through interactions between CD27 and CD70. Nat. Immunol. 6:412–418.

1524 NK CELL SUBSETS WITH DISTINCT FUNCTIONS

by guest on April 4, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from