DOI: 10.1126/scitranslmed.3006702, 208ra145 (2013);5 Sci Transl Med

et al.Amir HorowitzRevealed by Mass CytometryGenetic and Environmental Determinants of Human NK Cell Diversity

 Editor's Summary

   

tumors.contribute more to NK cell self-tolerance, whereas activating receptors may guide response to pathogens and mayexpression, activating receptors were controlled by the environment. These data suggest that inhibitory receptors

twins versus unrelated donors, they determined that although genetics primarily determined inhibitory receptordonors. They found up to 30,000 phenotypic NK cell populations in a given individual. What's more, by comparing the

The authors examined 35 parameters simultaneously in 5 sets of monozygotic twins as well as 12 unrelated

cytometry to examine NK cell diversity in humans. use masset al.combinations of activating and inhibiting receptors that govern their specificity. Now, Horowitz

exposure. However, this population is actually quite heterogeneous: Different subgroups of NK cells express different Natural killer (NK) cells were first discovered because of their ability to kill tumor cells without any previous

NK Cell Nature Versus Nurture

http://stm.sciencemag.org/content/5/208/208ra145.full.htmlcan be found at:

and other services, including high-resolution figures,A complete electronic version of this article

http://stm.sciencemag.org/content/suppl/2013/10/21/5.208.208ra145.DC1.html can be found in the online version of this article at: Supplementary Material

http://www.sciencemag.org/about/permissions.dtl in whole or in part can be found at: article

permission to reproduce this of this article or about obtaining reprintsInformation about obtaining

is a registered trademark of AAAS. Science Translational Medicinerights reserved. The title NW, Washington, DC 20005. Copyright 2013 by the American Association for the Advancement of Science; alllast week in December, by the American Association for the Advancement of Science, 1200 New York Avenue

(print ISSN 1946-6234; online ISSN 1946-6242) is published weekly, except theScience Translational Medicine

on

Nov

embe

r 25

, 201

3st

m.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Nov

embe

r 25

, 201

3st

m.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

R E S EARCH ART I C L E

IMMUNOLOGY

Genetic and Environmental Determinants of Human NKCell Diversity Revealed by Mass CytometryAmir Horowitz,1,2,3 Dara M. Strauss-Albee,3,4 Michael Leipold,2 Jessica Kubo,4

Neda Nemat-Gorgani,1 Ozge C. Dogan,4 Cornelia L. Dekker,5 Sally Mackey,5 Holden Maecker,2

Gary E. Swan,6 Mark M. Davis,2,3 Paul J. Norman,1 Lisbeth A. Guethlein,1 Manisha Desai,4

Peter Parham,1,2,3 Catherine A. Blish3,4*

Nov

embe

r 25

, 201

3

Natural killer (NK) cells play critical roles in immune defense and reproduction, yet remain the most poorlyunderstood major lymphocyte population. Because their activation is controlled by a variety of combinatoriallyexpressed activating and inhibitory receptors, NK cell diversity and function are closely linked. To provide anunprecedented understanding of NK cell repertoire diversity, we used mass cytometry to simultaneously ana-lyze 37 parameters, including 28 NK cell receptors, on peripheral blood NK cells from 5 sets of monozygotictwins and 12 unrelated donors of defined human leukocyte antigen (HLA) and killer cell immunoglobulin-likereceptor (KIR) genotype. This analysis revealed a remarkable degree of NK cell diversity, with an estimated 6000to 30,000 phenotypic populations within an individual and >100,000 phenotypes in the donor panel. Geneticslargely determined inhibitory receptor expression, whereas activation receptor expression was heavily environ-mentally influenced. Therefore, NK cells may maintain self-tolerance through strictly regulated expression ofinhibitory receptors while using adaptable expression patterns of activating and costimulatory receptors to respondto pathogens and tumors. These findings further suggest the possibility that discrete NK cell subpopulations couldbe harnessed for immunotherapeutic strategies in the settings of infection, reproduction, and transplantation.

on

stm

.sci

ence

mag

.org

Dow

nloa

ded

from

INTRODUCTION

Natural killer (NK) cells were discovered in the 1970s in both mice (1)and humans (1–3). They were first characterized by their ability torapidly kill tumor cells and later virus-infected cells (2, 4). Since thisinitial characterization, the nuanced complexity of NK cell phenotypesand function has been increasingly appreciated (5). The diversity ofthe NK cell repertoire is determined by the expression of an arrayof germline-encoded activating and inhibitory receptors (5–7). Thesereceptors interact with major histocompatibility complex class I andclass I–like molecules, costimulatory ligands, stress-related molecules,and cytokines (8–10). The combinatorial expression of this multitudeof receptors on an NK cell results in an abundance of mathematicallyfeasible subpopulations, but this diversity has yet to be captured insufficient detail.

Early studies using NK cell clones andmRNA expression data (11–13),more recent work incorporating monoclonal antibodies against a va-riety of NK cell receptors (14, 15), and bone marrow transplant datashowing the similarity of recipient to donor inhibitory repertoires (16)all demonstrate that host genetics are a critical determinant of inhib-itory receptor expression patterns. Additional nongenetically encodeddeterminants of diversity may influence activating receptor expression,including NK cell education, epigenetics, epistatic interactions betweenkiller cell immunoglobulin (Ig)–like receptor (KIR) genes, and viralinfections, particularly cytomegalovirus (CMV) (17–24). These obser-

1Department of Structural Biology, Stanford University School of Medicine, Stanford, CA94305, USA. 2Department of Microbiology and Immunology, Stanford University Schoolof Medicine, Stanford, CA 94305, USA. 3Stanford Immunology, Stanford University Schoolof Medicine, Stanford, CA 94305, USA. 4Department of Medicine, Stanford UniversitySchool of Medicine, Stanford, CA 94305, USA. 5Department of Pediatrics, StanfordUniversity School of Medicine, Stanford, CA 94305, USA. 6Center for Health Sciences, SRIInternational, Menlo Park, CA 94025, USA.*Corresponding author. E-mail: cblish@stanford.edu

www.Science

vations emphasize the need for a better understanding of the relation-ship between genotype, phenotype, and function in the formation ofthe NK cell repertoire.

Advances in fluorescence cytometry have laid the groundwork forunderstanding the NK cell repertoire, but the spectral limitations offluorescence have prevented exploration of its full extent. The recentdevelopment of mass cytometry, also known as cytometry by time offlight (CyTOF), provides an opportunity to overcome these limitations.Mass cytometry uses rare metal isotope-conjugated antibodies to simul-taneously detect up to 40 cellular markers. Because the metal isotopesare not naturally found in biological systems and have distinct time-of-flight profiles, background is nearly undetectable and the need forcompensation across channels is eliminated. Mass cytometry has beenused to profile human hematopoiesis (25) and to demonstrate the tre-mendous diversity within the CD8+ T cell compartment (26).

To provide a framework to better understand the NK cell reper-toire, we used mass cytometry in combination with high-resolutiongenotyping to evaluate the phenotypic heterogeneity of peripheralblood NK cells in 22 healthy individuals, including 5 sets of mono-zygotic twins. Our study shows a remarkable breadth and diversityin the human NK cell repertoire and the role of genetics and theenvironment in its formation and maintenance.

RESULTS

Mass cytometry for dissecting the phenotypic diversityof NK cellsTo better define the phenotypic diversity of human peripheral bloodNK cells, we designed a mass cytometry panel of 36 monoclonal anti-bodies recognizing lineage markers and NK receptors plus lanthanide-loaded DOTA-maleimide for dead cell discrimination (table S1). Because

TranslationalMedicine.org 23 October 2013 Vol 5 Issue 208 208ra145 1

R E S EARCH ART I C L E

this was the first mass cytometry panel specific for NK cell markers, wecompared results betweenmass cytometry and fluorescence cytometry.All antibodies successfully and comparably distinguished cell popula-tions on both platforms, with representative plots shown in fig. S1, Aand B. Staining with isotype control antibodies revealed minimal non-specific background by mass cytometry (fig. S1C). We also comparedNKcells stainedwith orwithout pretreatmentwith a humanFc receptorblocking solution and observed no significant differences in antibody

on

Nov

embe

r 25

, 201

3st

m.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

binding (fig. S1D). Adding to the versatilityand applicability of this approach, very sim-ilar NK cell staining profiles were observedfor fresh and cryopreserved peripheralblood mononuclear cell (PBMC) samples(fig. S1, E and F). NK cells identified witha serial gating strategy demonstrated arange of surface receptor expression levelsin different individuals (fig. S2).

Overall diversity of the NK cellreceptor phenotypeWith this mass cytometry platform, weinterrogated the diversity of the NK cellrepertoire in a cohort consisting of 12healthy, unrelated individuals and 5 setsof monozygotic twins of known humanleukocyte antigen (HLA) and KIR geno-type (Fig. 1 and fig. S3). After identifi-cation of NK cells with lineage markers,Boolean analysis was used to classify cellsas positive or negative for the expressionof each of 28 NK cell receptors (fig. S4).This allowed for the assessment of 228, or268,435,456, combinations of these recep-tors. The results revealed a much greaterheterogeneity in NK cell phenotypes thanhas been described previously (Fig. 1). Nosingle phenotype accounted for more than7% of the total NK cells, and only 14 of 28markers tested were expressed byNK cellsexhibiting the top 50 most frequent pheno-types (Fig. 1A). Furthermore, the subsetscomprising the top50phenotypes accountedfor an average of just 15% (range, 7 to 24%;SD, 4.8%) of an individual’s NK cells.

Although the conserved activating re-ceptors NKG2D and NKp46 are usuallydescribed as being expressed on mosthuman NK cells (27, 28), they were ex-pressed on just 28 and 25 of the top 50phenotypes, respectively. This low expres-sion is partially a result of our Booleangating strategy, in which only NK cellsthat were relatively bright for NKG2Dand NKp46 were considered positive (fig.S4). Thus, we may have underestimatedthe frequency of these markers, partic-ularly for cells that express low levels.Nonetheless, these results were concor-

www.Science

dant with the frequencies of NKG2D and NKp46 in total NK cell pop-ulations in this study, where NKG2D and NKp46 were highlyexpressed on an average of 48% (range, 29 to 67%) and 47% (range,28 to 71%) of NK cells, respectively, and with additional studies inwhich <100% of NK cells expressed these markers (27, 29–32). Theconserved inhibitory receptor NKG2A was more prevalent, foundon 40 of the top 50 phenotypes and always coexpressed with CD94,its partner polypeptide. In contrast with NKG2A, the highly variable

B C

A

0.01 0.1 1 100.01

0.1

1

10

Twin- 1

Twin- 3Twin- 4Twin-5

Twin- 2

% of cells

% o

f cel

ls

0.01 0.1 1 100.01

0.1

1

10

% of cells

% o

f cel

ls

6.8% 0%1%

Frequency of NK cells

CD127CD122CD572DS4

2DL2/L3/S22B4

CD94NKG2ANKG2DNKp30NKp46CD16

CD56dimCD56bright

HNK-001HNK-002HNK-003HNK-004HNK-005HNK-006HNK-007HNK-008HNK-009HNK-010HNK-011HNK-012Twin-1aTwin-1bTwin-2aTwin-2bTwin-3aTwin-3bTwin-4aTwin-4bTwin-5aTwin-5b

ρ = 0.65 ρ = 0.36

Fig. 1. Assessment of NK cell repertoire diversity based on expression of individual receptor pro-files. (A) Frequencies of the 50 most abundant NK cell phenotypes based on expression of 28 receptors

for 12 unrelated individuals (HNK-001 to HNK-012) and 5 pairs of monozygotic twins (Twin-1a/1b toTwin-5a/5b). Each column represents a phenotype, with colored boxes indicating receptor presence.(B) Comparative correlation of each twin with his/her identical twin of the frequencies of the 50 phe-notypes shown in (A). Spearman’s correlation coefficient is displayed. (C) Comparative correlation ofeach healthy unrelated individual with the other 11 individuals of the frequencies of the 50 phenotypesshown in (A).

TranslationalMedicine.org 23 October 2013 Vol 5 Issue 208 208ra145 2

R E S EARCH ART I C L E

KIRs contributed to relatively few phenotypes. KIR2DL1, KIR2DL5,and KIR3DL1 were not present in the 50 most frequent phenotypes.KIR2DL2/L3/S2, the most prevalent KIR group, was expressed on anaverage of 30% of an individual’s total NK cells, but on only 4 of the50 most frequent subpopulations, which together represented an av-erage of only 3.2% (range, 0.8 to 6.8%) of KIR2DL2/L3/S2-expressingcells. Thus, the remaining >93% on NK cells were distributed across arange of phenotypes, each represented by very few cells. These datareveal that a diverse spectrum of low-abundance NK cell phenotypesexpress inhibitory receptors.

To assess the genetic influences on the phenotypes of NK cells,we compared the frequencies of the top 50 NK cell phenotypes withinpaired monozygotic twins (Fig. 1B) and pairs of unrelated donors

on

Nov

embe

r 25

, 201

3st

m.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

(Fig. 1C). Twin pairs exhibited modestconcordance in the frequencies of NKcell subpopulations (Spearman’s r =0.65), although such concordance wasdiminished in pairs of unrelated donors(Spearman’s r = 0.36). These data suggestthat genetic differences between individ-uals may influence the repertoire of NKcell surface phenotypes.

NK cell receptor diversity in theexpression of inhibitory receptorsfor HLA class IWe next restricted the Boolean analysisto the six inhibitory NK cell receptorsKIR2DL1, KIR2DL2/L3/S2, KIR2DL5,KIR3DL1, NKG2A, and LILRB1, result-ing in 64 possible combinatorial cell sur-face phenotypes. Of these possibilities,the most frequently detected phenotypesin each individual included none of thesix inhibitory receptors or only NKG2A(Fig. 2A). These two subsets accounted foran average of 54% (range, 41 to 72%) of anindividual’s NK cells. Most remaining NKcells expressed one to three inhibitory re-ceptors. These results are consistent withthose obtained using fluorescence cytom-etry (15, 18, 33, 34).

In analyses parallel to those describedin Fig. 1 (B and C), we compared the expres-sion frequency of the inhibitory receptorcombinations between paired monozy-gotic twins (Fig. 2B) and between pairsof unrelated individuals (Fig. 2C). Concor-dance was extremely high within the pairsof monozygotic twins (Spearman’s r =0.87) and less marked in unrelated individ-uals (Spearman’s r = 0.77). These resultssuggest that host genetics determine theinhibitory receptor repertoire but onlyweakly affect the expression of other NKcell receptors (Fig. 1, B and C). Together,these data indicate that although the in-hibitory repertoire appears to be largely

www.Science

controlled by genetics, environmental and/or stochastic influences playa leading role in determining the overall repertoire of NK cell phenotypes.

Assessing the NK cell repertoire with clustering analysisBoolean analysis of the mass cytometry data resulted in thousandsof distinct NK cell phenotypes defined by combinatorial receptorexpression. However, because of the abundance of low-frequency pop-ulations, we used a clustering algorithm called spanning-tree pro-gression analysis of density-normalized events (SPADE) (25, 35). InSPADE analyses of PBMCs, nodes of phenotypically similar cellclusters corresponding to NK cells, CD4+ T cells, CD8+ T cells, B cells,and monocytes/macrophages were readily identified and distinguished(fig. S5).

B C

A

0.01 0.1 1 10 1000.01

0.1

1

10

100

% of cells

% o

f cel

ls

0.01 0.1 1 10 1000.01

0.1

1

10

100

Twin-1Twin-2Twin-3Twin-4Twin-5

% of cells

% o

f cel

ls

46% 0%1%Frequency of NK cells

2DL5LILRB12DL13DL1

2DL2/L3/S2NKG2A

HNK-001HNK-002HNK-003HNK-004HNK-005HNK-006HNK-007HNK-008HNK-009HNK-010HNK-011HNK-012Twin-1aTwin-1bTwin-2aTwin-2bTwin-3aTwin-3bTwin-4aTwin-4bTwin-5aTwin-5b

ρ = 0.87 ρ = 0.77

Fig. 2. Expression patterns of inhibitory NK cell receptors in unrelated individuals and mono-zygotic twins. (A) Inhibitory receptor profile of NK cells in 12 unrelated individuals and 5 pairs of mono-

zygotic twins as in Fig. 1A. The analysis was restricted to the evaluation of expression profiles of the sixinhibitory receptors KIR2DL1, KIR2DL2/L3/S2, KIR2DL5, KIR3DL1, LILRB1, and NKG2A. (B) Comparative cor-relation of each twin with his/her identical twin of the frequencies of the 50 phenotypes shown in (A).Spearman’s correlation coefficient is displayed. (C) Comparative correlation of each healthy unrelated in-dividual with the other 11 individuals of the frequencies of the 50 phenotypes shown in (A).

TranslationalMedicine.org 23 October 2013 Vol 5 Issue 208 208ra145 3

R E S EARCH ART I C L E

Consistent with the Boolean analysis, the distribution of inhibitoryreceptor expression on NK cells revealed by SPADE was more similarfor monozygotic twins (Fig. 3A and fig. S6) than for unrelated individ-uals (Fig. 3B and fig. S7). For instance, KIR2DL1 was observed onsimilar subpopulations in twins but not in unrelated individuals. Morediversity was observed in the intensity and frequency of expression ofactivating receptors, including NKG2D, which was expressed at vary-ing intensities and frequencies among different clusters between indi-viduals, even in twins. Additionally, HNK-003 and HNK-007 wereboth genotypically and phenotypically divergent for KIR2DS4,consistent with HNK-007 lacking a functional copy of the KIR2DS4gene (Fig. 3B and fig. S3). However, even the genotypically identical

www.Science

Twin-5a and Twin-5b showed differences in KIR2DS4-expressingnodes, with 28 KIRS2DS4-expressing nodes in common but 10 withdifferential expression. These data further illustrate the impact ofenvironmental factors on the total NK cell repertoire.

A comprehensive phenotype for CD57+NKG2C+

“memory-like” NK cellsSPADE analysis showed that five of the unrelated donors (HNK-002,HNK-004, HNK-005, HNK-007, and HNK-008) and one pair oftwins (Twin-5a and Twin-5b) had large NK cell subpopulations coex-pressing CD57 and NKG2C (Fig. 4A) that could represent memory-like NK cells generated in response to CMV infection (23, 24, 36, 37).

TranslationalMedicine.org 23

on

Nov

embe

r 25

, 201

3st

m.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

All of these subjects tested had positiveserology for CMV except for HNK-008(fig. S3). The remaining CMV-seropositivesubjects (HNK-001, HNK-012, Twin-4a,and Twin-4b) also had NKG2C+CD57+

NK cell subpopulations, although thesepopulations were less prominent. TheCD57+NKG2C+ cells expressed CD94 andCD16, but had low expression of CD122and KIR3DL1, and variable patterns ofexpression of KIR2DL1, KIR2DL3, NKG2D,NKp30, NKp46, and CD8 (Fig. 4, B andC). In the donors with only the HLA-C1epitope, NKG2C+CD57+NK cells prefer-entially expressed KIR2DL3, the inhibi-tory HLA-C1 receptor, whereas donorswith the HLA-C2 epitope preferentiallyexpressed KIR2DL1 (Fig. 4).

NK cell receptorcoexpression patternsWe next asked whether any single NKcell receptors were more likely to be ex-pressed with other single receptors onthe same NK cell. We therefore calculatedthe Spearman’s rank correlation of eachpossible receptor pair on each NK cellfrom all 22 donors (Fig. 5A). Weak neg-ative associations between the expressionof CD57 and NKp46, CD122, CD94, andNKG2A and weak positive associations be-tween the expression of NKG2A, CD94, andCD56 were the most dominant features.

Given the relative paucity of strongreceptor coexpression patterns, we per-formed hierarchical clustering of NK cellpopulations on the basis of surface recep-tor expression patterns to better understandNK cell population structure (Fig. 5B). Themarkers CD16, CD56, CD57, CD94, andNKG2A were distinct in their location atthe root of the tree, whereas most other NKcell receptors contributed equally to theoverall population structure (Fig. 5B). Fur-ther, principal components analyses in-dicated that NK cell populations expressing

A

5a

5b

5a

5b

003

007

003

007

B

Signal intensity

0 Max

£ 20£ 100£ 200£ 500≥ 1000

KIR2DL1

NKG2DNKG2A

KIR2DS4

KIR2DL1

NKG2DNKG2A

KIR2DS4

Fig. 3. Clustering analysis of NK cellsusing SPADE reveals diverse recep-tor expression patterns. (A) Repre-sentative SPADE trees of NKG2A, NKG2D,KIR2DL1, and KIR2DS4 for monozygotic

Twin-5a (top) and Twin-5b (bottom). Node color represents signal intensity of each marker, and size repre-sents frequency. (B) Representative SPADE trees of NKG2A, NKG2D, KIR2DL1, and KIR2DS4 for two healthyunrelated individuals: HNK-003 (top) and HNK-007 (bottom).

October 2013 Vol 5 Issue 208 208ra145 4

R E S EARCH ART I C L E

on

Nov

embe

r 25

, 201

3st

m.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

CD16, CD56, CD57, NKG2A, and CD94 were distinct and formed sepa-rate clusters (Fig. 5C). Overall, these results are consistent with a stochas-tic expression ofmost NK cell receptors, with lessmature NKG2A+CD94+

and more mature CD16+CD57+ cells representing the only majordistinction.

Quantifying the diversity of the NK cell repertoireHaving observed a stochastic expression of most NK receptors,leading to a broad range of possible combinatorial phenotypes on asingle cell, we next sought to quantify this high level of diversity.We reasoned that our system of NK cell phenotypes resembled a col-lection of species within a habitat, and adapted established ecologicalmethods. We first calculated the inverse Simpson diversity index,which measures the probability that two randomly selected organisms

www.ScienceTranslationalMedicine.org 23

from a population will be of the samespecies (38–40). This index accounts forboth the total abundance and the distri-bution of species within a population. Wefirst considered each NK cell as an indi-vidual organism, each NK cell phenotypeas a species, and the group of phenotypesexpressing each cell surface marker as apopulation. Inverse Simpson indices foreach population illustrate the contributionof each marker’s cellular distribution to adonor’s total phenotypic diversity (Fig. 6A).

Receptors that were highly expressedon NK cells, such as CD56, CD16, CD94,and 2B4, had high average inverse Simpsonindices, indicating their contribution tonumerous NK cell phenotypes (Fig. 6A).Strikingly, we also observed a broad rangewithin the cohort in the diversity of theNK cell populations expressing these andother receptors. For example, although theaverage inverse Simpson index of CD56-expressing populations was 659, the mini-mumwas 135 and themaximumwas 1017,yielding a range of 882.We expected high-ly polymorphic KIRs to exhibit broadranges of inverse Simpson indices, whichwas true forKIR3DL1(range, 844),KIR2DS4(range, 582), and KIR2DL1 (range, 460).However, many nonpolymorphic or min-imally polymorphic receptors such asCD16 (range, 938) and 2B4 (range, 691)showed similar or even greater rangesof inverse Simpson indices. Thus, we ob-served a high degree of variability amongdonors in the diversity of the NK cell pop-ulations on which specific receptors wereexpressed, which was not wholly dependenton allelic polymorphism.

We next asked whether host geneticsplayed any role in determining this vari-ability by calculating the correlation be-tween the inverse Simpson indices for eachreceptor in each pair of twins versus all

unrelated individuals. Despite the broad range in receptor diversityacross individuals, the inverse Simpson indices for most receptorswithin twin pairs were remarkably similar, with high intraclass corre-lation coefficients (>0.6) for all receptors except CD56, 2B4, CD127,CD117, LILRB1, HLA-DR, and CCR7 (Fig. 6B). This suggests thathost genetics may be a determinant of the range of NK cell pheno-types on which a given receptor is expressed, with factors in additionto receptor polymorphisms exerting considerable influence on eachreceptor’s diversity.

Next, to understand the range of NK cell diversity in a human pop-ulation, we calculated the inverse Simpson indices for each individualdonor, rather than each receptor (Fig. 6C). Eighteen of the 22 individualshad total inverse Simpson indices in the relatively restricted range of 350to 900, suggestive of an optimal level of human NK cell diversity.

Fig. 4. Identification and phenotypic evaluation of CD57+NKG2C+ memory-like NK cells. (A) Skele-ton structures of a SPADE tree from healthy unrelated individuals (HNK) and monozygotic twins (Twin)

with NKG2C-expressing nodes highlighted with red shading. (B) Representative SPADE trees of theNKG2C-expressing nodes shown in (A) from five unrelated individuals. (C) Representative SPADE treesof NKG2C-expressing nodes for one pair of monozygotic twins, as in (B).

October 2013 Vol 5 Issue 208 208ra145 5

R E S EARCH ART I C L E

on

Nov

embe

r 25

, 201

3st

m.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

We next asked how many total NK cell phenotypes were likely tobe present in each individual. We used the Chao 1 nonparametric spe-cies estimator, which projects the total number of species present in asample by using the abundance of rare species as a predictor of un-discovered species (38). This method predicted between 6000 and30,000 distinct NK cell phenotypes within each individual, with a

www.Science

median of 15,000 (Fig. 6D). In this analysis, twins showed a low in-traclass correlation of 0.278 (Fig. 6B), suggesting that the combina-torial assortment of receptors in distinct NK cell phenotypes maybe more a function of environmental than genetic influences.

Calculating the expanse of the NK cell repertoireFinally, we calculated how many NK cell phenotypes were likely to bepresent at a population level. Because our experimental analyses of3500 to 35,000 NK cells per donor were unlikely to have capturedthe total NK cell diversity, we generated a sample-based rarefactioncurve to provide an estimate of the total number of NK cell pheno-types as a function of the number of phenotypes sampled (Fig. 7A).This analysis estimated the expansiveness of this cohort’s repertoire at124,651 NK cell phenotypes. We also predicted the total number ofNK cell species on a population level using three other independentand well-established ecological methods (38). Together, despite theirvariable estimation approaches and input parameters, all of these cal-culations yielded remarkably similar estimates of 108,000 to 125,000total NK cell phenotypes (Fig. 7B).

DISCUSSION

Using multiparametric mass cytometry to examine expression of 28NK cell surface markers, we achieved an incisive dissection of humanNK cell repertoires. A vast phenotypic diversity was uncovered, whichis influenced by the genetic differences between individual humansas well as by the environmental differences they experience. Geneticdifferences strongly influence the combinatorial expression patternsof the inhibitory receptors that recognize HLA class I. This reper-toire of inhibitory receptors, which is required for maintenance of self-tolerance as well as for NK cell “education” or “licensing,” is overlaidby an environmentally influenced diversity in the expression of acti-vating and costimulatory receptors. This complementary combinationresults in NK cell repertoires that comprise between 6000 and 30,000phenotypically distinguishable subpopulations. Beyond such diversitywithin an individual is further variation between individuals. Morethan 100,000 NK cell subpopulations were distinguished in the small“population” of 22 peoplewe studied. These data reveal a potentialmech-anism by which NK cells could maintain self-tolerance through strictlyregulated expression of inhibitory receptors while using more malleableand adaptable expression patterns of activating and costimulatory re-ceptors to respond to infections and cancers.

Comparing adult monozygotic twins distinguished the effects ofgenetic and environmental factors. The NK cells from twins had verysimilar patterns of expression of inhibitory HLA class I receptors,consistent with previous comparisons of unrelated individuals show-ing that the inhibitory receptor repertoire is influenced by variabilityin the KIR and HLA class I genes (1, 11, 12, 15, 18, 33). Although verysimilar overall, the inhibitory receptor repertoires of twins exhibitedsome differences, which might be explained by environmental influ-ences, including well-established epigenetic alterations of the KIR genes(18–20).

We observed extensive diversity within populations of cells ex-pressing each combination of inhibitory receptors. For instance, theonly other marker consistently coexpressed on NKG2A-expressingcells is CD94, its partner polypeptide. Even coexpression of markersbelieved to be expressed on all or most NK cells, such as NKp46 and

Fig. 5. NK cell receptor coexpression patterns and population struc-ture. (A) Spearman’s rank correlation matrices of coexpression profiles for

each receptor pair for all NK cells from all 22 donors. (B) Hierarchicalclustering dendrogram of NK cell receptor profiles from all NK cells fromall 22 donors. (C) Principal components analysis of receptor expressionpatterns from all NK cells from all 22 donors. Dots indicate the contribu-tions of each receptor to principal component 1 (PC1) (x axis) versus prin-cipal component 2 (PC2) (y axis).

TranslationalMedicine.org 23 October 2013 Vol 5 Issue 208 208ra145 6

R E S EARCH ART I C L E

on

Nov

embe

r 25

, 201

3st

m.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

NKG2D, was highly variable. Thus, NKG2A is not expressed by oneor a few subpopulations of NK cells, but by thousands. Because bothNKG2A and its ligand, HLA-E, are highly conserved (6, 41), this di-versity may provide an additional degree of flexibility to the repertoire.

Associated with the paucity of pairwise combinations of receptorsin the overall NK cell population (Fig. 5) is our finding that the ex-pression pattern for most of the receptors (including inhibitory KIRs,the C-type lectin receptors, and natural cytotoxicity receptors) is deter-mined solely by the genes encoding the receptor (Fig. 6). This suggestsa predetermined level for both the expression and distribution of thesereceptors by NK cells. Several other receptors—HLA-DR, CCR7, CD127,CD117, 2B4, and LILRB1—had diversity patterns more suggestive ofenvironmental influence. These receptors are known to vary in responseto activation [HLA-DR and CCR7 (8, 42)], differentiation [CD117 (43)],infection [2B4 and LILRB1 (14, 44)], and pregnancy [LILRB1 (45)].

Variation in a receptor’s expression within an individual could beexplained by genetic polymorphisms that affect expression level (18–20).In particular, KIR polymorphisms greatly influence the avidity and

www.ScienceTranslationalMedicine.org 23

specificity of these interactions, leadingto a highly variable repertoire (17, 46).Bolstering this argument is the broad rangeof expression by KIR3DL1, the most di-verse of all KIR (47), and KIR2DS4 (48, 49).CD16, a less polymorphic receptor, is themost diverse of the NK cell receptors inits expression across cellular subpopula-tions. One interpretation of this findingis that a diversity of NK cells facilitatesthe critical function of CD16 in mediat-ing antibody-dependent cellular cytotox-icity (ADCC), a bridge between innateand adaptive immunity. Alternatively, thediversity of CD16 could reflect its role asthe only receptor demonstrated to be suf-ficient for NK cell activation (50).

“Null” cells, which express none of thecell surface markers other than CD56, arethe most frequent NK cell subpopulation.They could represent a transitional statecapable of up-regulating activating or in-hibitory receptors. If so, they may be lesslikely to up-regulate expression of inhib-itory receptors, because the inhibitory re-ceptor expression patterns are extremelystable over time (51). However, even cellswith no inhibitory receptors may play afunctional role. Although these “unedu-cated” cells lacking receptors for self-HLAare hyporesponsive to HLA-deficient tar-get cells (52), uneducated cells can be ef-fective in ADCC (53).

Our results suggest that the human NKcell repertoire is anchored by the combina-tion of a mature CD16/CD57-expressingpopulation and a less mature NKG2A/CD94-expressing population. Stochasticassortment of additional receptors createstremendous additional diversity within

this framework. Expansions of NKG2C+CD57+ NK cells that pref-erentially expressed KIR reactive with self–HLA-C molecules wereidentified in several healthy, mostly CMV-seropositive, donors. Thesememory-like cells had a mature phenotype, as previously reported(22, 54). Despite this subset’s presumed expansion in response to asingle pathogen, CMV (17), these memory-like NK cells had highlyvariable expression levels of other markers, further underlining theoverall diversity of the NK cell population.

Several limitations of the study should be noted. First, our samplesize was modest and serial sampling was not performed, which limitedour ability to compare NK cell phenotypes between individuals ofsimilar KIR/HLA genotype. In addition, the number of NK cells sam-pled per donor was between ~3000 and ~35,000, which is well belowthe theoretical number of 268,435,456 potential phenotypes, limitingour sampling of rare phenotypes. Finally, by using a Boolean gatingstrategy in which we defined cell populations as positive for a givenreceptor only if high levels were expressed, we may have underesti-mated the frequency of some receptors.

0 150 300 450 600 750 900 10500

2

4

6

8

Inverse Simpson index bin center

Nu

mb

er o

f in

div

idu

als

1 2 3 4 5 6 7 8 9 10 11 12 1a 1b 2a 2b 3a 3b 4a 4b 5a 5b0

10,000

20,000

30,000

HNK Twins

CD56CD94

CD162B4

CD57

2DL2/L3/S

2

NKG2D

NKp30

CD122

CD127

NKp462DL1

3DL1

NKG2A2DS4

NKG2CCD8

CD117

NKp44

LILRB1

HLA-DR

CD4CD25

CD27

CCR72DL5

2DL4

TRAIL0

500

1000

Inve

rse

Sim

pso

nIn

dex

Twin-3b

Twin-3a

Unrelated donor

All donors

A B

C D

VariableIntraclass correlation coefficient

CD56 0.068CD16 0.815CD94 0.8882B4 0.24

CD122 0.887CD57 0.772

NKp30 0.937KIR2DL2/L3/S2 0.824

NKG2D 0.893NKp46 0.819CD127 0.315

KIR2DL1 0.605NKG2A 0.945

KIR3DL1 0.923KIR2DS4 0.874NKG2C 0.846

CD8 0.605CD117 0.222CD27 0.844

NKp44 0.75CD4 0.801

LILRB1 0.175HLA-DR 0.224

CD25 0.736CCR7 –0.816

KIR2DL5 0.86KIR2DL4 0.648

TRAIL 0.803

Chao 1 estimator

0.278

Ch

ao 1

pre

dic

ted

ph

eno

typ

es

Fig. 6. Genetics dictate the expression of most single NK cell receptors, but not the combinatorialassortment of phenotypes. (A) Inverse Simpson for the population of cells expressing each single re-

ceptor. Box plots show the median, 25th and 75th percentiles, and total range of all 22 donors. Yellowlines show a representative twin pair. Red line indicates an unpaired representative twin. Red and yellowcolors correspond to matching donors in (D). (B) Summary of intraclass correlation coefficients of theinverse Simpson index of twin pairs versus all 22 individuals. (C) Histogram of inverse Simpson indicesfor the total NK cell population of each donor. (D) Total phenotypes predicted by the Chao 1 non-parametric species estimator (calculated as described in Materials and Methods) for all donors.

October 2013 Vol 5 Issue 208 208ra145 7

R E S EARCH ART I C L E

on

Nov

embe

r 25

, 201

3st

m.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

NK cells are quick, versatile lymphocytes that function in innateimmunity, adaptive immunity, and reproduction. In performing thesevital functions, we now see how each human individual has an almostincredible diversity of NK cell subsets. Furthermore, the repertoire ofthese subsets differs greatly from one person, or patient, to another.Among these different subsets are developmental intermediates inpathways of maturation and differentiation, as well as different typesof effector NK cell. The latter include effector and memory NK cells,which exhibit, either alone or in combination, some measure of spec-ificity for a pathogen, tumor, or allogeneic tissue. The immediate chal-lenge for research is to define the physiological contributions of thedifferent NK cell subsets. The future challenge will be using that

www.Science

knowledge to design specific, effective NK cell–mediated therapiesagainst infectious, malignant, and reproductive diseases.

MATERIALS AND METHODS

Study designThe objective of this study was to profile the baseline diversity of thehealthy human NK cell repertoire, providing a framework to under-stand the roles of NK cells in health and in disease pathogenesis. Tothis end, we performed an observational study of 22 healthy individ-uals, including 5 pairs of monozygotic twins, a sample size that wasdetermined both by feasibility and to allow us to sample a represent-ative pool of donors of different KIR and HLA genotypes. This was anobservational study; randomization and blinding were not used. Allphenotypic analyses were performed ex vivo and not run in replicates.Analyses of twins versus unrelated individuals allowed us to examinethe influence of host genetics versus environment on the NK cellrepertoire.

Human subjects and cellsThis study was performed in accordance with the Declaration ofHelsinki and approved by the Institutional Review Boards of StanfordUniversity and SRI International. Cryopreserved PBMCs were ob-tained from three sources. First, 10 generally healthy individuals (5 male,5 female; aged 24 to 60 years; median age, 34 years) were recruited atStanford University. Second, PBMCs from two healthy anonymous do-nors were obtained through purchase from the Stanford Blood Center.Third, PBMC samples from five sets of monozygotic twin pairs (fourmale, six female; aged 20 to 40 years; median age, 23 years) were col-lected as part of an ongoing study of influenza vaccination involvingparticipants in the Twin Research Registry at SRI International. Fulldetails of the twin registry have been described (55). All subjects gavefully informed and written consent.

Staining, data acquisition, and analysisPBMCs were thawed and washed with RPMI 1640 (Corning Cellgro)with 10% heat-inactivated fetal bovine serum, 2 mM L-glutamine, andantibiotics [penicillin (100 U/ml) and streptomycin (100 mg/ml)] (GibcoBRL/Life Technologies) and rested at 37°C with 5% CO2 for 4 hours.Two million PBMCs were stained for mass cytometry analyses as de-scribed (25, 26) with the antibodies listed in table S1, and data wereacquired on a CyTOF instrument (DVS Sciences). Fluorescence cy-tometry analyses were acquired on an LSR II (BD Biosciences) withthe antibodies in table S2. FCS files were analyzed with FlowJosoftware v9.4.8 (Tree Star Inc.). Boolean analyses were performed witha custom Python script (http://www.python.org) to sort cells on thebasis of positive or negative expression of each marker. SPADE analy-ses were performed as described (35).

Antibody conjugationAntibodies were purchased from companies specified in table S1 andlabeled withMaxPar X8 labeling reagent kits (DVS Sciences) accordingto the manufacturer’s instructions.

KIR and HLA genotyping and CMV serologyKIR gene presence and HLA-A,-B,-C alleles were determined by poly-merase chain reaction–based sequence-specific oligonucleotide probe

A

B

0 2 106 4 106 6 106 8 106 1 1070

25,000

50,000

75,000

100,000

125,000

# NK cells included

Ph

eno

typ

e co

un

t

Smax = 124,651 unique phenotypes

Rarefaction curve

Phenotype counts based on HNKs 1-22

Rarefaction Second-order

Jackknife

Chao 1 Chao 20

50,000

100,000

150,000

Pre

dic

ted

ph

eno

typ

es

Fig. 7. Calculating the expanse of the NK cell repertoire. (A) Sample-based rarefaction predicting the total number of NK cell phenotypes within

the cohort of 22 donors. The number of phenotypes observed wascalculated for incremental inputs of 50,000 NK cells up to and includingthe total 304,909 NK cells sampled in all 22 donors, reshuffling the orderof donor sampling each of 10 times. These data were used to plot a curve,which, when extrapolated to an asymptote, provided an estimate of thetotal NK cell population. (B) Nonparametric species estimators confirmquantification of the total number of NK cell phenotypes.

TranslationalMedicine.org 23 October 2013 Vol 5 Issue 208 208ra145 8

R E S EARCH ART I C L E

on

Nov

embe

r 25

, 201

3st

m.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

with a Luminex 100 instrument (Luminex Corp.). The assays wereperformed with LABType SSO reagents (One Lambda). KIR genotypingat allele-level resolution was determined by pyrosequencing as described(56). DNA from Twin-1 was not available for typing, but monozygositywas previously established (55). CMV serology was obtained with theanti-CMV IgG ELISA (enzyme-linked immunosorbent assay) kit (GoldStandard Diagnostics) according to the manufacturer’s suggestions.Plasma samples for CMV testing from the healthy unrelated individ-uals were obtained about 1 year after PBMC sampling.

Quantification of diversityThe inverse Simpson index was calculated with the following equation:

D ¼ 1

∑S

i¼1p2i

where S is the total number of species, and pi is the proportionalabundance of the ith species.

Sample-based rarefactionMean species occurrences of the number of phenotypes observed wereplotted, and the curve was projected by fitting the following asymp-totic equation:

SðnÞ ¼ Smaxn

Bþ n

where n is the number of species included, Smax is the total numberof species projected, and B is a fitted constant.

Nonparametric species estimatorsThe Chao 1 estimator is given by the following equation:

S1 ¼ Sobs þ F21

2F2

where Sobs is the number of species observed, F1 is the number of spe-cies observed in exactly one cell, and F2 is the number observed inexactly two cells.

The Chao 2 estimator, which incorporates occurrence rather thanabundance data, is given by the following equation:

S2 ¼ Sobs þ Q21

2Q2

where Q1 is the number of species observed in exactly one donor, andQ2 is the number observed in exactly two donors.

The second-order Jackknife estimator, which reduces samplingbias via calculation with a subset of the data removed, is given bythe following equation:

Sjack2 ¼ Sobs þ Q1ð2m − 3Þm

−Q2ðm − 2Þ2mðm − 1Þ

" #

where m is the total number of samples.

Statistical analysisStatistical analyses were performed with Excel (Microsoft Corp.), STATAv10 (Stata Corp.), Prism v5 (GraphPad Software Inc.), and the open-source statistical package R (http://www.r-project.org).

www.Science

SUPPLEMENTARY MATERIALS

www.sciencetranslationalmedicine.org/cgi/content/full/5/208/208ra145/DC1Fig. S1. Validation of mass cytometry panel for phenotyping NK cells.Fig. S2. Gating strategy for defining NK cells and ranges of frequencies of NK cells expressingeach individual marker.Fig. S3. KIR and HLA class I genotyping.Fig. S4. Representative two-dimensional mass cytometry plots of NK cells from one represent-ative donor (HNK-002) showing expression of all 28 NK cell receptors individually whenexpressed against CD56.Fig. S5. SPADE trees showing distribution of cell lineage markers across unfractionated PBMC.Fig. S6. Clustering analysis of NK cells from Twin-5a and Twin-5b using SPADE.Fig. S7. Clustering analysis of NK cells from HNK-003 and HNK-007 using SPADE.Table S1. Mass cytometry panel antibody information.Table S2. Mass cytometry isotype control and LSR antibody panel information.

REFERENCES AND NOTES

1. R. Kiessling, E. Klein, H. Wigzell, “Natural” killer cells in the mouse. I. Cytotoxic cells withspecificity for mouse Moloney leukemia cells. Specificity and distribution according togenotype. Eur. J. Immunol. 5, 112–117 (1975).

2. R. R. Herberman, J. R. Ortaldo, G. D. Bonnard, Augmentation by interferon of humannatural and antibody-dependent cell-mediated cytotoxicity. Nature 277, 221–223 (1979).

3. J. R. Ortaldo, G. D. Bonnard, P. D. Kind, R. B. Herberman, Cytotoxicity by cultured humanlymphocytes: Characteristics of effector cells and specificity of cytotoxicity. J. Immunol.122, 1489–1494 (1979).

4. C. A. Biron, R. M. Welsh, Activation and role of natural killer cells in virus infections.Med. Microbiol. Immunol. 170, 155–172 (1982).

5. J. P. Di Santo, Functionally distinct NK-cell subsets: Developmental origins and biologicalimplications. Eur. J. Immunol. 38, 2948–2951 (2008).

6. M. Colonna, Specificity and function of immunoglobulin superfamily NK cell inhibitory andstimulatory receptors. Immunol. Rev. 155, 127–133 (1997).

7. D. H. Raulet, W. Held, I. Correa, J. R. Dorfman, M. F. Wu, L. Corral, Specificity, tolerance anddevelopmental regulation of natural killer cells defined by expression of class I-specificLy49 receptors. Immunol. Rev. 155, 41–52 (1997).

8. L. L. Lanier, Natural killer cells: From no receptors to too many. Immunity 6, 371–378 (1997).9. H. G. H. Ljunggren, K. K. Kärre, In search of the “missing self ”: MHC molecules and NK cell

recognition. Immunol. Today 11, 237–244 (1990).10. E. Vivier, S. Ugolini, Natural killer cells: From basic research to treatments. Front. Immunol. 2,

18 (2011).11. N. M. N. Valiante, M. M. Uhrberg, H. G. H. Shilling, K. K. Lienert-Weidenbach, K. L. K. Arnett,

A. A. D’Andrea, J. H. J. Phillips, L. L. L. Lanier, P. P. Parham, Functionally and structurallydistinct NK cell receptor repertoires in the peripheral blood of two human donors. Immunity 7,739–751 (1997).

12. N. N. Wagtmann, S. S. Rajagopalan, C. C. C. Winter, M. M. Peruzzi, E. O. E. Long, Killer cellinhibitory receptors specific for HLA-C and HLA-B identified by direct binding and by func-tional transfer. Immunity 3, 801–809 (1995).

13. M. Colonna, J. Samaridis, Cloning of immunoglobulin-superfamily members associated withHLA-C and HLA-B recognition by human natural killer cells. Science 268, 405–408 (1995).

14. M. Sternberg-Simon, P. Brodin, Y. Pickman, B. Onfelt, K. Kärre, K. J. Malmberg, P. Höglund,R. Mehr, Natural killer cell inhibitory receptor expression in humans and mice: A closerlook. Front. Immunol. 4, 1–12 (2013).

15. M. Yawata, N. Yawata, M. Draghi, F. Partheniou, A. M. Little, P. Parham, MHC class I–specificinhibitory receptors and their ligands structure diverse human NK-cell repertoires towarda balance of missing self-response. Blood 112, 2369–2380 (2008).

16. H. G. Shilling, K. L. McQueen, N. W. Cheng, J. A. Shizuru, R. S. Negrin, P. Parham, Reconstitutionof NK cell receptor repertoire following HLA-matched hematopoietic cell transplantation.Blood 101, 3730–3740 (2003).

17. V. Béziat, L. Liu, J. A. Malmberg, M. A. Ivarsson, E. Sohlberg, A. T. Björklund, C. Retière,E. Sverremark-Ekström, J. Traherne, P. Ljungman, M. Schaffer, D. A. Price, J. Trowsdale,J. Michaëlsson, H. G. Ljunggren, K. J. Malmberg, NK cell responses to cytomegalovirusinfection lead to stable imprints in the human KIR repertoire and involve activating KIRs.Blood 121, 2678–2688 (2013).

18. K. Schönberg, M. Sribar, J. Enczmann, J. C. Fischer, M. Uhrberg, Analyses of HLA-C–specificKIR repertoires in donors with group A and B haplotypes suggest a ligand-instructedmodel of NK cell receptor acquisition. Blood 117, 98–107 (2011).

19. K. Schönberg, J. C. Fischer, G. Kögler, M. Uhrberg, Neonatal NK-cell repertoires are functionally,but not structurally, biased toward recognition of self HLA class I. Blood 117, 5152–5156 (2011).

TranslationalMedicine.org 23 October 2013 Vol 5 Issue 208 208ra145 9

R E S EARCH ART I C L E

on

Nov

embe

r 25

, 201

3st

m.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

20. F. Cichocki, J. S. Miller, S. K. Anderson, Y. T. Bryceson, Epigenetic regulation of NK celldifferentiation and effector functions. Front. Immunol. 4, 55 (2013).

21. N. K. Bjorkström, T. Lindgren, M. Stoltz, C. Fauriat, M. Braun, M. Evander, J. Michaëlsson,K. J. Malmberg, J. Klingström, C. Ahlm, H. G. Ljunggren, Rapid expansion and long-termpersistence of elevated NK cell numbers in humans infected with hantavirus. J. Exp. Med. 208,13–21 (2011).

22. C. Petitdemange, P. Becquart, N. Wauquier, V. Béziat, P. Debré, E. M. Leroy, V. Vieillard,W. M. Yokoyama, Unconventional repertoire profile is imprinted during acute chikungunyainfection for natural killer cells polarization toward cytotoxicity. PLoS Pathog. 7, e1002268(2011).

23. S. Lopez-Vergès, J. M. Milush, B. S. Schwartz, M. J. Pando, J. Jarjoura, V. A. York, J. P. Houchins,S. Miller, S. M. Kang, P. J. Norris, D. F. Nixon, L. L. Lanier, Expansion of a unique CD57+NKG2Chi

natural killer cell subset during acute human cytomegalovirus infection. Proc. Natl. Acad. Sci.U.S.A. 108, 14725–14732 (2011).

24. B. Foley, S. Cooley, M. R. Verneris, M. Pitt, J. Curtsinger, X. Luo, S. Lopez-Vergès, L. L. Lanier,D. Weisdorf, J. S. Miller, Cytomegalovirus reactivation after allogeneic transplantation pro-motes a lasting increase in educated NKG2C+ natural killer cells with potent function.Blood 119, 2665–2674 (2012).

25. S. C. Bendall, E. F. Simonds, P. Qiu, E. A. D. Amir, P. O. Krutzik, R. Finck, R. V. Bruggner,R. Melamed, A. Trejo, O. I. Ornatsky, R. S. Balderas, S. K. Plevritis, K. Sachs, D. Pe’er, S. D. Tanner,G. P. Nolan, Single-cell mass cytometry of differential immune and drug responses across ahuman hematopoietic continuum. Science 332, 687–696 (2011).

26. E. W. E. Newell, N. N. Sigal, S. C. S. Bendall, G. P. G. Nolan, M. M. M. Davis, Cytometry bytime-of-flight shows combinatorial cytokine expression and virus-specific cell niches within acontinuum of CD8+ T cell phenotypes. Immunity 36, 142–152 (2012).

27. S. S. Bauer, V. V. Groh, J. J. Wu, A. A. Steinle, J. H. J. Phillips, L. L. L. Lanier, T. T. Spies,Activation of NK cells and T cells by NKG2D, a receptor for stress-inducible MICA. Science285, 727–729 (1999).

28. S. Sivori, M. Falco, L. Moretta, A. Moretta, Extending killer Ig-like receptor function: FromHLA class I recognition to sensors of microbial products. Trends Immunol. 31, 289–294(2010).

29. A. Silva, D. M. Andrews, A. G. Brooks, M. J. Smyth, Y. Hayakawa, Application of CD27 as amarker for distinguishing human NK cell subsets. Int. Immunol. 20, 625–630 (2008).

30. M. Della Chiesa, M. Falco, M. Podestà, F. Locatelli, L. Moretta, F. Frassoni, A. Moretta, Phenotypicand functional heterogeneity of human NK cells developing after umbilical cord blood trans-plantation: A role for human cytomegalovirus? Blood 119, 399–410 (2012).

31. P. André, R. Castriconi, M. Espéli, N. Anfossi, T. Juarez, S. Huë, H. Conway, F. Romagné, A. Dondero,M. Nanni, S. Caillat-Zucman, D. H. Raulet, C. Bottino, E. Vivier, A. Moretta, P. Paul, Comparativeanalysis of human NK cell activation induced by NKG2D and natural cytotoxicity receptors.Eur. J. Immunol. 34, 961–971 (2004).

32. J. M. Milush, B. R. Long, J. E. Snyder-Cappione, A. J. Cappione, V. A. York, L. C. Ndhlovu,L. L. Lanier, J. Michaëlsson, D. F. Nixon, Functionally distinct subsets of human NK cellsand monocyte/DC-like cells identified by coexpression of CD56, CD7, and CD4. Blood114, 4823–4831 (2009).

33. N. K. Björkström, P. Riese, F. Heuts, S. Andersson, C. Fauriat, M. A. Ivarsson, A. T. Björklund,M. Flodström-Tullberg, J. Michaëlsson, M. E. Rottenberg, C. A. Guzmán, H. G. Ljunggren,K. J. Malmberg, Expression patterns of NKG2A, KIR, and CD57 define a process ofCD56dim NK-cell differentiation uncoupled from NK-cell education. Blood 116, 3853–3864(2010).

34. M. Yawata, N. Yawata, M. Draghi, A. M. Little, F. Partheniou, P. Parham, Roles for HLA andKIR polymorphisms in natural killer cell repertoire selection and modulation of effectorfunction. J. Exp. Med. 203, 633–645 (2006).

35. P. P. Qiu, E. F. E. Simonds, S. C. S. Bendall, K. D. K. Gibbs, R. V. R. Bruggner, M. D. M. Linderman,K. K. Sachs, G. P. G. Nolan, S. K. S. Plevritis, Extracting a cellular hierarchy from high-dimensionalcytometry data with SPADE. Nat. Biotechnol. 29, 886–891 (2011).

36. M. Gumá, A. Angulo, C. Vilches, N. Gómez-Lozano, N. Malats, M. López-Botet, Imprint ofhuman cytomegalovirus infection on the NK cell receptor repertoire. Blood 104, 3664–3671(2004).

37. M. Gumá, M. Budt, A. Sáez, T. Brckalo, H. Hengel, A. Angulo, M. López-Botet, Expansion ofCD94/NKG2C+ NK cells in response to human cytomegalovirus-infected fibroblasts. Blood107, 3624–3631 (2006).

38. N. J. Gotelli, R. K. Colwell, Estimating species richness, in Biological Diversity: Frontiers inMeasurement and Assessment, A. E. Magurran, B. J. McGill, Eds. (Oxford Univ. Press, Oxford,2010), pp. 39–54.

39. N. J. Gotelli, R. K. Colwell, Quantifying biodiversity: Procedures and pitfalls in the measurementand comparison of species richness. Ecol. Lett. 4, 379–391 (2001).

40. R. K. Colwell, Biodiversity: Concepts, patterns, and measurement, in The Princeton Guide toEcology (Princeton Univ. Press, Princeton, 2009), pp. 257–263.

41. B. P. Shum, L. R. Flodin, D. G. Muir, R. Rajalingam, S. I. Khakoo, S. Cleland, L. A. Guethlein,M. Uhrberg, P. Parham, Conservation and variation in human and common chimpanzeeCD94 and NKG2 genes. J. Immunol. 168, 240–252 (2002).

www.ScienceT

42. J. H. Evans, A. Horowitz, M. Mehrabi, E. L. Wise, J. E. Pease, E. M. Riley, D. M. Davis, A distinctsubset of human NK cells expressing HLA-DR expand in response to IL-2 and can aid immuneresponses to BCG. Eur. J. Immunol. 41, 1924–1933 (2011).

43. B. Grzywacz, N. Kataria, M. Sikora, R. A. Oostendorp, E. A. Dzierzak, B. R. Blazar, J. S. Miller,M. R. Verneris, Coordinated acquisition of inhibitory and activating receptors and func-tional properties by developing human natural killer cells. Blood 108, 3824–3833(2006).

44. S. Jost, J. Reardon, E. Peterson, D. Poole, R. Bosch, G. Alter, M. Altfeld, Expansion of 2B4+

natural killer (NK) cells and decrease in NKp46+ NK cells in response to influenza. Immunology132, 516–526 (2011).

45. R. Marlin, M. Duriez, N. Berkane, C. De Truchis, Y. Madec, M. A. Rey-Cuille, J. S. Cummings,C. Cannou, H. Quillay, F. Barré-Sinoussi, M. T. Nugeyre, E. Menu, Dynamic shift fromCD85j/ILT-2 to NKG2D NK receptor expression pattern on human decidual NK during the firsttrimester of pregnancy. PLoS One 7, e30017 (2012).

46. H. G. Hilton, L. Vago, A. M. Older Aguilar, A. K. Moesta, T. Graef, L. Abi-Rached, P. J. Norman,L. A. Guethlein, K. Fleischhauer, P. Parham, Mutation at positively selected positions in thebinding site for HLA-C shows that KIR2DL1 is a more refined but less adaptable NK cellreceptor than KIR2DL3. J. Immunol. 189, 1418–1430 (2012).

47. P. J. Norman, L. Abi-Rached, K. Gendzekhadze, D. Korbel, M. Gleimer, D. Rowley, D. Bruno,C. V. F. Carrington, D. Chandanayingyong, Y. H. Chang, C. Crespí, G. Saruhan-Direskeneli,P. A. Fraser, K. Hameed, G. Kamkamidze, K. A. Koram, Z. Layrisse, N. Matamoros, J. Milà,M. H. Park, R. M. Pitchappan, D. D. Ramdath, M. Y. Shiau, H. A. F. Stephens, S. Struik, D. H. Verity,R. W. Vaughan, D. Tyan, R. W. Davis, E. M. Riley, M. Ronaghi, P. Parham, Unusual selec-tion on the KIR3DL1/S1 natural killer cell receptor in Africans. Nat. Genet. 39, 1092–1099(2007).

48. T. Graef, A. K. Moesta, P. J. Norman, L. Abi-Rached, L. Vago, A. M. Older Aguilar, M. Gleimer,J. A. Hammond, L. A. Guethlein, D. A. Bushnell, P. J. Robinson, P. Parham, KIR2DS4 is aproduct of gene conversion with KIR3DL2 that introduced specificity for HLA-A*11 whilediminishing avidity for HLA-C. J. Exp. Med. 206, 2557–2572 (2009).

49. K.-J. Malmberg, J. Michaëlsson, P. Parham, H.-G. Ljunggren, Killer cell immunoglobulin-likereceptor workshop: Insights into evolution, genetics, function, and translation. Immunity35, 653–657 (2011).

50. Y. T. Bryceson, M. E. March, H. G. Ljunggren, E. O. Long, Synergy among receptors onresting NK cells for the activation of natural cytotoxicity and cytokine secretion. Blood107, 159–166 (2006).

51. H. G. H. Shilling, N. N. Young, L. A. L. Guethlein, N. W. N. Cheng, C. M. C. Gardiner, D. D. Tyan,P. P. Parham, Genetic control of human NK cell repertoire. J. Immunol. 169, 239–247(2002).

52. N. Anfossi, P. André, S. Guia, C. S. Falk, S. Roetynck, C. A. Stewart, V. Breso, C. Frassati, D. Reviron,D. Middleton, F. Romagné, S. Ugolini, E. Vivier, Human NK cell education by inhibitory recep-tors for MHC class I. Immunity 25, 331–342 (2006).

53. N. Tarek, J. B. Le Luduec, M. M. Gallagher, J. Zheng, J. M. Venstrom, E. Chamberlain,S. Modak, G. Heller, B. Dupont, N. K. V. Cheung, K. C. Hsu, Unlicensed NK cells targetneuroblastoma following anti-GD2 antibody treatment. J. Clin. Invest. 122, 3260–3270(2012).

54. M. R. Goodier, C. M. Mela, A. Steel, B. Gazzard, M. Bower, F. Gotch, NKG2C+ NK cells areenriched in AIDS patients with advanced-stage Kaposi’s sarcoma. J. Virol. 81, 430–433(2006).

55. R. E. Krasnow, L. M. Jack, C. N. Lessov-Schlaggar, A. W. Bergen, G. E. Swan, The Twin ResearchRegistry at SRI International. Twin Res. Hum. Genet. 16, 463–470 (2013).

56. P. J. Norman, L. Abi-Rached, K. Gendzekhadze, J. A. Hammond, A. K. Moesta, D. Sharma,T. Graef, K. L. McQueen, L. A. Guethlein, C. V. F. Carrington, D. Chandanayingyong, Y. H. Chang,C. Crespí, G. Saruhan-Direskeneli, K. Hameed, G. Kamkamidze, K. A. Koram, Z. Layrisse,N. Matamoros, J. Milà, M. H. Park, R. M. Pitchappan, D. D. Ramdath, M. Y. Shiau, H. A. F. Stephens,S. Struik, D. Tyan, D. H. Verity, R. W. Vaughan, R. W. Davis, P. A. Fraser, E. M. Riley, M. Ronaghi,P. Parham, Meiotic recombination generates rich diversity in NK cell receptor genes, alleles,and haplotypes. Genome Res. 19, 757–769 (2009).

Acknowledgments: We thank E. Newell, E. Simonds, A. Moudgil, J. Haddon, H. Hilton, andP. Bollyky for helpful discussions and/or critical reading of the manuscript. Twins were re-cruited from the Twin Research Registry at SRI International; we wish to thank M. McElroy,L. Jack, R. Krasnow, J. Rubin, D. Basin, L. Panini, and M. Ritchey, and funds from SRI’s Centerfor Health Sciences and NIH grants DA011170, DA023063, AI057229, AI090019, and ES022153.We also thank the twins for their contributions to science. We thank the clinical staff of theStanford–Lucile Packard Children’s Hospital Vaccine Program including S. Swope, RN, C. Walsh,RN, A. Trela, RN, phlebotomist M. Ugur, and clinical research assistants A. Goel, K. Spann, R. Fleischmann,S. Batra, and I. Chang. Funding: This study was funded by NIH training grant T32 AI07290 (A.H.),NSF training grant DGE-114740 (D.M.S.-A.), NIH grants AI22039 (P.P.) and U19AI090019 andU19 AI057229 (M.M.D.), NIH/National Center for Research Resources Clinical and TranslationalScience Award UL1 RR025744, a New Investigator Award from the University of WashingtonCenter for AIDS Research (P30 AI027757, C.A.B.), and a Beckman Young Investigator Award (C.A.B.).

ranslationalMedicine.org 23 October 2013 Vol 5 Issue 208 208ra145 10

R E S EARCH ART I C L E

Author contributions: A.H. and C.A.B. designed and performed the experiments, analyzed thedata, and wrote the manuscript; D.M.S.-A. analyzed the data, performed ecological analyses,and wrote the manuscript; M.L., N.N.-G., O.C.D., H.M., P.J.N., and L.A.G. performed the experimentsand contributed to the writing of the manuscript; C.A.B., J.K., and M.D. performed statisticalanalyses; G.E.S. referred twin pairs for recruitment; C.L.D. and S.M. conducted the clinical trialto provide samples; G.E.S., C.L.D., and M.M.D. contributed to the writing of the manuscript; A.H.,P.P., and C.A.B. conceived the experiments and wrote the manuscript. Competing interests: Theauthors declare that they have no competing interests. Data and materials availability: Alldata are deposited in ImmPort with access number SDY232.

www.ScienceT

Submitted 4 June 2013Accepted 19 August 2013Published 23 October 201310.1126/scitranslmed.3006702

Citation: A. Horowitz, D. M. Strauss-Albee, M. Leipold, J. Kubo, N. Nemat-Gorgani, O. C. Dogan,C. L. Dekker, S. Mackey, H. Maecker, G. E. Swan, M. M. Davis, P. J. Norman, L. A. Guethlein,M. Desai, P. Parham, C. A. Blish, Genetic and environmental determinants of human NK celldiversity revealed by mass cytometry. Sci. Transl. Med. 5, 208ra145 (2013).

ranslationalMedicine.org 23 October 2013 Vol 5 Issue 208 208ra145 11

on

Nov

embe

r 25

, 201

3st

m.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om