Genetic Control of Human NK Cell Repertoire

Transcription

1 This information is current as of December 29, Genetic Control of Human NK Cell Repertoire Heather G. Shilling, Neil Young, Lisbeth A. Guethlein, Nathalie W. Cheng, Clair M. Gardiner, Dolly Tyan and Peter Parham J Immunol 2002; 169: ; ; doi: /jimmunol References Subscription Permissions Alerts This article cites 55 articles, 20 of which you can access for free at: Why The JI? Submit online. Rapid Reviews! 30 days* from submission to initial decision No Triage! Every submission reviewed by practicing scientists Fast Publication! 4 weeks from acceptance to publication *average Information about subscribing to The Journal of Immunology is online at: Submit copyright permission requests at: Receive free -alerts when new articles cite this article. Sign up at: Downloaded from by guest on December 29, 2018 The Journal of Immunology is published twice each month by The American Association of Immunologists, Inc., 1451 Rockville Pike, Suite 650, Rockville, MD Copyright 2002 by The American Association of Immunologists All rights reserved. Print ISSN: Online ISSN:

2 The Journal of Immunology Genetic Control of Human NK Cell Repertoire 1 Heather G. Shilling,* Neil Young, 2 * Lisbeth A. Guethlein,* Nathalie W. Cheng,* Clair M. Gardiner, 3 * Dolly Tyan, and Peter Parham 4 * Through differential killer cell Ig-like receptor (KIR) and CD94:NKG2 gene expression, human NK cells generate diverse repertoires, each cell having an inhibitory receptor for autologous HLA class I. Using a new method for measuring repertoire difference that integrates multiple flow cytometry parameters, we found individual repertoire stability, but population variability. Correlating repertoire differences with KIR and HLA genotype for 85 sibling pairs reveals the dominant influence of KIR genotype; HLA genotype having a subtle, modulating effect on relative KIR expression frequencies. HLA and/or KIR genotype also influences CD94:NKG2A expression. After HLA-matched stem cell transplantation, KIR repertoires either recapitulated that of the donor or were generally depressed for KIR expression. Human NK cell repertoires are defined by combinations of variable KIR and HLA class I genes and conserved CD94:NKG2 genes. The Journal of Immunology, 2002, 169: Natural killer cells are lymphocytes of innate immunity that kill cells and produce cytokines (1). NK cells contribute to early defense against viral infection and have been implicated in alloreactions following bone marrow transplantation (1 3). NK cell clonal diversity is due to the combinatorial expression of different cell surface molecules, including several receptors for polymorphic epitopes of MHC class I, each encoded by a conventional, nonrearranging gene. NK cells can express receptors for allogeneic as well as autologous MHC molecules, and class I specificity of individual clones is determined by the array of receptors present at the cell surface, rather than any individual element (4, 5). In humans, the family of killer cell Ig-like receptors (KIR) 5 provides inhibitory, and some activating, receptors for HLA-A, -B, -C, and -G (5). KIR are not present in mice, where analogous functions are performed by the Ly49 family of lectinlike receptors (6). KIR are encoded by a compact, rapidly evolving family of genes in the leukocyte receptor complex on human chromosome 19, which exhibits extensive haplotypic variation in gene number and content, and allelic polymorphism for individual genes (7 10). Complementing KIR, the CD94:NKG2 lectin-like molecules provide inhibitory and activating NK cell receptors specific for *Departments of Structural Biology, and Microbiology and Immunology, Stanford University School of Medicine, Stanford, CA 94305; and Department of Pediatrics, Cedars-Sinai Medical Center, Los Angeles, CA Received for publication February 19, Accepted for publication April 19, The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 This research was supported by National Institutes of Health Grants AI22039, AI45865, and PO1 CA H.G.S. and L.A.G. were supported by National Institutes of Health Training Grant AI07290 and C.M.G. by a Leukemia Research Foundation Fellowship. Sample collection at City of Hope was supported by National Institutes of Health Grant PO1 CA Current address: Nuffield Department of Surgery, University of Oxford, John Radcliffe Hospital, Oxford, U.K. OX3 9DU. 3 Current address: Department of Biochemistry, Trinity College Dublin, Dublin 2, Ireland. 4 Address correspondence and reprint requests to Dr. Peter Parham, Department of Structural Biology, Sherman Fairchild Building, 299 Campus Drive West, Stanford University School of Medicine, Stanford, CA address: peropa@leland.stanford.edu 5 Abbreviations used in this paper: KIR, killer cell Ig-like receptor; CML, chronic myelogenous leukemia; MUD, matched unrelated donor; GVHD, graft-vs-host disease. complexes of HLA-E with peptides derived from HLA class I leader sequences (11 13). Contrasting with KIR, genes encoding CD94 and NKG2 family members are conserved in the human population (14); phylogenetically they are also more conserved, being present in mouse and linked to Ly49 genes in the NK complex (15 17). The human NK complex on chromosome 12 contains genes for CD94 and NKG2 family members, as well as a nonfunctional gene, Ly49L; the only human homolog of the mouse Ly49 gene family (18 22). A general rule is that NK cells cannot kill cells expressing a full complement of autologous MHC class I allotypes, but can kill cells expressing some combinations of allogeneic MHC class I (23). In humans, this tolerance to self has been correlated with the expression of at least one inhibitory KIR or CD94:NKG2A receptor having specificity for self-hla class I (4). Such observations, derived from functional assays, provided evidence that HLA class I type influences the repertoire of inhibitory HLA class I receptors expressed by peripheral blood NK cells. In contrast, population analysis of HLA-B- or -C-specific KIR expression revealed no difference between individuals who did, or did not, express a cognate HLA class I ligand (24, 25). Conversely, one study reporting no HLA effect implicated undefined, non-hla genes as factors affecting NK cell expression of the HLA-B-specific KIR (KIR3DL1) (24). Subsequent discovery of KIR population diversity made the KIR locus on chromosome 19 a candidate for the non-hla genes (7, 26 31). The investigation described here tested this hypothesis and defined the magnitude of the role of HLA class I in NK cell repertoire selection. In this analysis, the confounding effects of KIR and HLA genetic diversity were controlled by comparing NK cell receptor repertoires in sibling pairs of known HLA and KIR types. The role of the KIR genotype in determining KIR repertoire was further assessed by following KIR expression after HLA-matched allogeneic and autologous stem cell transplantation. The results demonstrate a role for both KIR and HLA class I genes in determining human NK cell repertoires and resolve the seemingly paradoxical results of previous functional and genetic analyses. Materials and Methods Healthy sibling donors Peripheral blood samples were obtained from 104 healthy individuals: 53 donors representing 17 families, of which 5 were multigenerational, from Copyright 2002 by The American Association of Immunologists, Inc /02/$02.00

3 240 GENETIC CONTROL OF HUMAN NK CELL REPERTOIRE Cedars-Sinai Medical Center (Los Angeles, CA); 4 individuals representing 2 families from the City of Hope Histocompatibility Laboratory (Duarte, CA); and from 47 individuals representing 17 families from Stanford Medical Center Histocompatibility Laboratory (Stanford, CA). Stem cell patients and donors Eighteen patients receiving allogeneic stem cell transplants at the Stanford University Medical Center for treatment of chronic myelogenous leukemia (CML; n 12) or acute myelogenous leukemia (n 6) were studied. Of these, 12 received bone marrow (n 8) or G-CSF-mobilized peripheral blood stem cells (n 4) from HLA-matched sibling donors. The remaining six patients received bone marrow from unrelated donors (MUD). This group included 6 males and 12 females, with a median age of 39 years (range, years). Pretransplant myeloablative regimens included a combination of busulfan, cyclophosphamide, and etoposide, with or without fractionated total-body irradiation. Posttransplant graft-vs-host disease (GVHD), antiviral, and antibacterial prophylaxis were comparable (32 35). Blood was drawn from donors and recipients before transplant and from patients upon clinical engraftment or by day 30 posttransplant. Subsequent sample collection targeted days 30, 60, 90, and/or 100, 120, 150, 180, 270, and 1 year posttransplant. Five non-hodgkin s lymphoma patients undergoing autologous peripheral blood stem cell transplantation at Stanford University Medical Center were studied. This group included three males and two females, with a mean age of 54 years (range, years). Patients underwent standard G-CSF stem cell mobilization and harvest and comparable myeloablative conditioning (36). Samples were collected before transplant and on the first day that each patient showed clinical engraftment or by day 14 posttransplant. Samples were collected weekly for the first month following transplant and on days 60, 90, and 180 posttransplant. PBMC were isolated by Ficoll-Hypaque gradient separation. All samples were collected with approval of the appropriate Institutional Review Board. KIR nomenclature KIR2DL1, KIR2DL3, KIR3DL1, and KIR3DL2 alleles were named according to guidelines used in naming HLA alleles. Briefly, an asterisk separates the accepted gene designations (37) from three digits which distinguish alleles that differ by nonsynonymous substitutions; fourth and fifth digits were assigned to alleles that differ only by synonymous substitutions. Numerical order was assigned based on date of submission to GenBank; partial sequences and splice variants were excluded, as were sequences of single PCR-derived clones. Genomic DNA and cdna preparation Genomic DNA was prepared from PBMC using a QIAamp Blood kit (Qiagen, Valencia, CA). Total cellular RNA was prepared from PBMC using RNAzol (Tel-Test, Friendswood, TX). First-strand cdna was synthesized from 5.0 g RNA using oligo(dt) (PE Applied Biosystems, Foster City, CA) and avian myeloblastosis virus/reverse transcriptase (Promega, Madison, WI) at 42 C for 90 min. HLA class I determination and KIR epitope-typing HLA-A and HLA-B Ags, including Bw4 and Bw6, were determined serologically by the laboratories supplying the sample. HLA-C type was determined serologically or by PCR-sequence-specific primer analysis of genomic DNA using C locus Sequence-Specific Primer Unitray test kits (Pel-Freez Biologicals, Rogers, AR). The presence of the class I HLA-C KIR epitopes was determined by RT-PCR amplification. In addition, the HLA-Bw4 and Bw6 serological typing results were confirmed by RT-PCR. The group 1 HLA-C epitope was detected using specific sense primer (5 -CGA GTG AGC CTG CGG AAC-3 ) plus an HLA-C locus-specific antisense primer (5 -AGG ACA GCT AGG ACA RCC-3 ); the group 2 HLA-C epitope was detected with a specific sense primer (5 -CCG AGT GAA CCT GCG GAA A-3 ) and the HLA-C generic primer. HLA-Bw4 epitopes were detected with a mixture of three sense primers (5 -CCT GCG CAC CGC GCT CC-3, 5 -CCT GCG GAT CGC GCT CC-3, and 5 -CCT GCG GAC CCT GCT CC-3 ) with an HLA-B locus-specific antisense primer (5 -TCC GAT GAC CAC AAC TGC T-3 ). HLA-Bw6 epitopes were detected with a specific sense primer (5 -CCT GCG GAA CCT GCG CG-3 ) paired with the same generic antisense primer. Primers were used at 0.5 M in25- l reactions using 1 2 l cdna. Internal control primers specific for -actin (sense, 5 -CTT CGA GCA AGA GAT GGC CAC-3 ; antisense, 5 -TTG CTG ATC CAC ATC TGC TGG AAG-3 ) were included at 0.05 M. PCR conditions were: initial denaturation at 95 C for 5 min; 5 cycles of 97 C for 20 s, 62 C for 45 s, and 72 C for 90 s; followed by cycles of 95 C for 20 s, 60 C for 45 s, and 72 C for 90 s; and a 7-min extension at 72 C. KIR genotyping Typing of genomic DNA for 10 KIR was performed as described by Uhrberg et al. (7), with modification. KIR2DL1v (KIR2DL1*004) was detected with the KIR2DL1 forward primer and KIR2DS1 reverse primer. Detection of KIR2DS5 was as modified by Vilches et al. (38). For KIR subtyping, primers designed to discriminate allele-specific polymorphisms were paired with KIR2DL1, KIR2DL3, KIR3DL1, or KIR3DL2 locus-specific primers. KIR3DL1 and KIR3DL2 subtyping were as described by Gardiner et al. (29) and Shilling et al. (39); KIR2DL1 and KIR2DL3 subtyping were as described by Shilling et al. (39). To supplement this, KIR2DL1, KIR2DL3, and KIR3DL1 locus-specific PCR products from some sibling pairs were purified using a QIAquick PCR Purification kit (Qiagen) and partially sequenced with the original amplification primers by a dye terminator automated sequencer (Applied Biosystems). Abs and flow cytometric analysis mabs DX27 (anti-kir2dl2/kir2dl3/kir2ds2), DX9 (anti-kir3dl1), DX31 (anti-kir3dl2; generously provided by L. Lanier, Cancer Research Center, University of California, San Francisco, CA), EB6 (anti-kir2dl1/ 2DS1), and Z199 (anti-nkg2a) (Beckman-Coulter-Immunotech, Brea, CA) were each used in combination with anti-cd3 (SK7) and anti-cd56 (NCAM16.2; BD Biosciences, Mountain View, CA) in individual threecolor flow cytometry assays of PBMC from each donor. After gating on the CD3 CD56 lymphocyte (i.e., NK cell) subset, KIR- or CD94:NKG2Aexpressing populations were identified using irrelevant, isotype-matched control Ab stains to set the lower limit of the KIR- or CD94:NKG2Apositive gate, thereby excluding background fluorescence and receptornegative cells. Frequency and median fluorescence intensity of NK cells binding each specific Ab were calculated for these populations. For 9 of the 104 donors surveyed, CD94:NKG2A expression was determined by gating on NK cells which stained brightly with a CD94-specific Ab (HP-3D9; BD Biosciences) instead of Z199. Anti-CD3 and anti-cd56 Abs were labeled with PerCP and FITC, respectively, anti-kir, anti-nkg2a, and anti-cd94 Abs were PE conjugated. Flow cytometric analysis was performed on a FACScan flow cytometer using CellQuest software (BD Biosciences). The four KIR-specific Abs used here detect all KIR of known specificity, thereby providing a measure of functional KIR expression. As KIR-specific Abs bind subsets of KIR, rather than individual allotypes, fluorescence intensity levels likely encompass various Ab-binding affinities as well as cell surface expression levels, and therefore reflect physical properties of KIR repertoire. Calculation of differences in frequency and median fluorescence intensity of KIR expression from flow cytometry data The difference in frequency of KIR expression by NK cells for each KIRspecific Ab stain was calculated according to the formula: (frequency (sample 1) frequency (sample 2))/(mean frequency (samples 1 and 2)) difference in frequency. The mean was incorporated as a denominator here to amplify differences between KIR-negative and KIR-positive samples. The four frequency differences were added to determine the sum of frequency differences. Differences in median fluorescence intensity were calculated using the same formula, then added to give the sum of median fluorescence intensity differences. Statistical calculations Correlations coefficients (r) were calculated using the formula r Cov(X,Y)/ X Y, where Cov(X,Y) (X X )(Y Y ); significance was evaluated by two-tailed t test with n 2 df. Results NK cell KIR repertoire is stable within an individual but varies between individuals Human NK cell clones differ in number and combination of KIR expressed (4, 40). This mode of gene expression generates a diverse repertoire of KIR expression in peripheral blood NK cells. To assess the stability of these expression patterns, we studied peripheral blood NK cells of five healthy donors over a period of about 1 year, analyzing cell surface KIR expression with four

4 The Journal of Immunology mabs of different KIR specificity. Flow cytometry analysis showed that the proportion of NK cells binding each Ab and their median level of binding were stable over time for all five donors (Fig. 1, A E). These parameters appeared not to be perturbed by infection or other environmental stress, including running of a marathon by the donor in Fig. 1A. The results showed that donors have stable and characteristic patterns of KIR expression, or KIR repertoires, which can be described in terms of eight flow cytometry measurements: two parameters for each of four Abs. To compare KIR repertoires, paired sets of flow cytometry data of the type shown in Fig. 1, A E, were used to calculate summed differences for the proportion of NK cells binding each Ab and for the median level of Ab binding. In this way, a comparison between two data sets was reduced to a single point on a two-dimensional plot (Fig. 1F; see Materials and Methods for further details). Control comparisons confirmed the validity of this method; when frozen aliquots of the same sample of PBMC were thawed and analyzed in duplicate, and, on different days, the total differences were small (data not shown). As expected from Fig. 1, A E, comparison FIGURE 1. Patterns of KIR expression by peripheral blood NK cells can be used to define and compare NK cell KIR repertoires. A E, Flow cytometry data for five donors whose peripheral blood NK cells were assayed at intervals over a period of 13 mo for binding for KIR-specific Abs. Each panel shows the data from one donor. The Abs encompass the HLA-A, -B, and -C-specific KIR and were EB6 (E; KIR2DL1/S1), DX27 (u; KIR2DL2 3/S2), DX9 ( ; KIR3DL1), and DX31 (Œ; KIR3DL2). Each identical symbol in a panel represents a different time point. F, The data from A E is used to compare KIR repertoires. Each symbol on the two-dimensional plot derives from two sets of flow cytometry data and provides a measure of the overall differences in frequency of cells binding each Ab and the overall differences in the level of binding. Comparisons are for pairs of unrelated individuals ( ) and, as controls, for data sets obtained from the same person, but from blood samples drawn on different occasions ( ). The range of possible value for each sum is of the repertoires measured from PBMC obtained from the same donor on different occasions produced small differences. In contrast, pairwise comparison of unrelated individuals revealed a wide range of differences (Fig. 1F). In sum, these results show that an individual s NK cell KIR repertoire is defined and stable over time, but that repertoires are highly diversified within human populations. Differences in KIR repertoire are principally determined by KIR genotype Functional and genetic studies have given contradictory results regarding the role of HLA class I polymorphism in determining NK cell KIR repertoires, while the effects of KIR gene variation on NK cell KIR repertoire remain unexplored. To define the contributions of these two gene families, we compared the KIR repertoires of 85 healthy sibling pairs and correlated differences in KIR expression with identity or disparity at the HLA class I and KIR loci. A panel of 104 individuals from 36 families was studied (Fig. 2). Of 85 sibling pairs, 21 (25%) were HLA class I identical and 64 (75%) were disparate at one or both HLA class I haplotypes, as determined by serology (HLA-A and -B) and/or PCR typing (HLA-C). KIR identity of sibling pairs was determined through a combination of low- and high-resolution PCR typing and selected DNA sequencing; results were confirmed by segregation for 43 of the 85 sibling pairs. Twenty-seven sibling pairs (32%) were KIR identical and 58 (68%) were disparate at one or both KIR haplotypes. These numbers demonstrate random segregation of parental HLA and KIR haplotypes. That the number of KIR-identical sibling pairs exceeds 25% is because some families segregate more than one copy of a common KIR haplotype. HLA class I and KIR polymorphisms independently segregated in the 36 families, as expected from location of HLA and KIR genes to different chromosomes (chromosomes 6 and 19, respectively). Seven pairs (8%) were KIR and HLA identical; 20 (24%) were KIR identical, HLA disparate; 14 (16%) were KIR disparate, HLA identical; and 44 (52%) were KIR and HLA disparate (Fig. 2). Peripheral blood NK cells were analyzed for KIR expression by flow cytometry (as in Fig. 1, A E). For each sibling pair, the difference in NK cell KIR repertoire was calculated (as in Fig. 1F), the results being displayed in four two-dimensional dot plots in Fig. 3 according to whether the sibling pairs were KIR identical, HLA identical (Fig. 3A); KIR identical, HLA disparate (Fig. 3B); KIR disparate, HLA identical (Fig. 3C); or KIR disparate, HLA disparate (Fig. 3D). Each dot plot has a distinct distribution, demonstrating the influence of KIR and HLA genes on the NK cell KIR repertoire. However, KIR type is by far the dominant factor. In the context of incompatibility at the other genetic complex, KIR identity gave greater similarity in repertoire than did HLA class I identity. Thus, siblings who are KIR identical, HLA disparate have a much tighter distribution with lower summed difference (Fig. 3B) than do siblings who are KIR disparate, HLA identical (Fig. 3C). The dominance of KIR over HLA is well illustrated when the data points for all sibling pairs are placed on the same dot plot and differentiated according to either KIR (Fig. 3E) orhla identity (Fig. 3F). The data points for all KIR-identical pairs cluster close to the origin, irrespective of their HLA status (Fig. 3E), whereas the data points for HLA-identical pairs distribute throughout all but the outer fringe of the distribution (Fig. 3F). The distributions for KIRdisparate sibling pairs split into two subpopulations (Fig. 3, C F), where the subpopulation with higher difference consists of pairs in which one sibling (but not the other) lacks reactivity with a KIR Ab (DX9 or EB6), due either to the absence of a KIR gene (KIR3DL1 or KIR2DL1/2DS1) or presence of an allele giving a

5 242 GENETIC CONTROL OF HUMAN NK CELL REPERTOIRE FIGURE 2. KIR genotyping of 85 HLA-typed sibling pairs. The sibling pairs are organized into four groups according to genotype: KIR-identical, HLA-identical (A), KIR-identical, HLA-disparate (B), KIR-disparate, HLA-identical (C) and KIR-disparate, HLA-disparate (D). Siblings having matching HLA-A, -B, and -C allotypes by serology (HLA-A and -B) and/or PCR typing (HLA-C) were considered to be HLA identical. Siblings with matching KIR typing and allelic subtyping results were considered KIR identical, with the exception of one sibling pair (no. 48), found to be KIR disparate by partial sequencing of KIR locus-specific PCR products. The presence of KIR genes is indicated with white boxes, with allelic subtypes shown only for those sibling pairs having matching combinations of KIR genes. Dark gray boxes indicate KIR genes that are absent. In heterozygous individuals, allele names are separated with a slash. Alleles which are not discriminated by subtyping are listed together, e.g., as KIR2DL3*002 or 6; KIR3DL1*002 refers here to a set of similar alleles which includes KIR3DL1*002, *003, and * null phenotype (KIR3DL1*004) with DX9 (29). Thus, KIR genes dominate over HLA class I genes in determining the NK cell repertoire of KIR expression in peripheral blood NK cells. Apparent from the distribution of repertoire differences between sibling pairs is that the two summed differences are not completely independent variables. Thus, siblings who have large differences in Ab-binding frequencies tend to have large differences in Ab-binding levels (Fig. 3E). This correlation between the sum of differences in frequency and median levels of Ab binding is statistically significant (r 0.72, p 0.01; Fig. 3E). It demonstrates that polymorphisms in the KIR gene family influence the frequency of NK cells that express a particular KIR as well as other characteristics assessed by Ab binding, including the level of cell surface expression and Ab-binding affinity. HLA genotype influences the frequencies of KIR-expressing NK cells To determine the contribution of HLA class I to the NK cell KIR repertoire, we considered just the subset of KIR-identical sibling pairs. For these 27 pairs the differences in repertoire between siblings in HLA-identical pairs were compared with the differences between the siblings in HLA-disparate pairs. As seen in Fig. 4, the 7 HLA-identical pairs have similarly low values for the sum of differences for the frequency of Ab-binding cells, whereas the 20 HLA-disparate KIR-identical siblings exhibited a much wider range of values. This difference has statistical significance in a t test ( p ). In contrast, the HLA-identical and HLA-disparate pairs exhibited a similar range of values for the summed

6 The Journal of Immunology 243 FIGURE 3. KIR genes dominate HLA class I genes in determining human peripheral blood NK cell KIR repertoires. Sibling NK cell KIR repertoires were compared for 85 HLA- and KIR-typed sibling pairs. Sibling pairs are organized into four groups: KIR identical, HLA identical (A), KIR identical, HLA disparate (B), KIR disparate, HLA identical (C) and KIR disparate, HLA disparate (D). Each symbol (F) represents the repertoire difference between two siblings. E and F, Both contain all of the data points from A D. E, Sibling pairs are distinguished according to KIR identity (f) or difference ( ); F, they are distinguished according to HLA identity ( ) or difference ( ). Included in E is a linear regression trend line, calculated using all data, which shows the correlation between summed differences in frequency and surface level of KIR expression. differences in the median level of Ab binding. Thus, the effect of the HLA class I genotype on the NK cell KIR repertoire is to modify the relative frequencies of cells expressing particular KIR, but not the surface levels of KIR expression. Best characterized of the HLA class I-specific inhibitory KIR are those specific for epitopes of HLA-B or -C, which correlate with polymorphisms in the carboxyl terminal part of the 1 helix. The HLA-Bw4 epitope, formed by some sequence motifs at positions in HLA-B, is recognized by KIR3DL1. HLA-C epitopes, defined by alternative motifs at positions 77 and 80, are bound by KIR2DL2/2DL3/2DS2 and 2DL1/2DS1, respectively (41 44). Because different HLA types can have the same KIR epitopes, we distinguished KIR-identical, HLA-disparate sibling pairs according to whether they were identical or different for KIR epitopes (Fig. 4). Repertoire differences were similar, there being no statistically significant difference between the two HLA-disparate groups. When the expression of individual KIR was analyzed for these sibling pairs, neither KIR frequency nor fluorescence intensity differences between siblings were consistently correlated with the presence or absence of a relevant KIR epitope. However, FIGURE 4. HLA class I influences the relative frequencies of cells expressing KIR but not their levels of surface expression. Shown are KIR repertoire comparisons for the 27 KIR-identical sibling pairs. They are distinguished as HLA identical (f), HLA different but having identical HLA-B and -C KIR epitopes (u) orhla and KIR epitope different ( ). KIR- and HLA-identical sibling pairs and KIR-identical, HLA-disparate, KIR epitope-matched pairs were significantly different ( p 0.01). Likewise, KIR- and HLA-identical pairs were different from KIR-identical, HLA-disparate, KIR epitope-disparate pairs ( p 0.001). Together, these results imply that HLA modification of the NK cell KIR repertoire cannot be explained simply in terms of the HLA-B and -C sequence motifs. CD94:NKG2A expression is affected by HLA and KIR genotypes The stability of CD94:NKG2A expression was studied using the same five donors and samples used for KIR (Fig. 1). For each individual, the frequency of NK cells expressing CD94:NKG2A and their median level of surface expression were stable over time. However, these parameters varied among the five donors (data not shown). The stability of the individual CD94:NKG2A phenotype parallels that seen for KIR. CD94:NKG2 expression by NK cells from the 85 sibling pairs was measured. A range of difference between siblings, in both frequency and level of CD94:NKG2A expression, was observed (Fig. 5). The differences were least for KIR- and HLA-identical siblings (Fig. 5A), showing that the combination of KIR and HLA influences the repertoire of CD94:NKG2A expression. The differences were lower for KIR-disparate, HLA-identical siblings than for either KIR-identical, HLA-disparate siblings or siblings disparate at both KIR and HLA. Thus, HLA type appears to override KIR type in affecting CD94:NKG2 expression. However, the effect is weaker than the influences of HLA and KIR on KIR expression and for this sample size did not reach statistical significance. We also compared expression of CD94:NKG2A and KIR in the donor panel, independently of familial relationships. This revealed a statistically significant inverse correlation between the total frequency of cells expressing KIR and the frequency of cells expressing CD94:NKG2. This was seen either when the total frequency of KIR-expressing cells was roughly estimated by summation of the frequencies of KIR expression for the four Abs (Fig. 5E, r 0.39 ( p 0.01)) or when it was calculated from individual KIR frequencies, taking account of knowledge that KIR are expressed in random combinations (4) (Fig. 5F, r 0.40 ( p 0.01), respectively). This result demonstrates some coupling between expression of the CD94:NKG2A receptor and KIR receptors.

7 244 GENETIC CONTROL OF HUMAN NK CELL REPERTOIRE FIGURE 5. CD94:NKG2A expression by NK cells is influenced by HLA class I and KIR type. Differences in frequency of cells expressing CD94:NKG2A and in level of expression are shown for the 85 sibling pairs. As for KIR repertoire comparisons, CD94:NKG2A expression differences were calculated by dividing the absolute difference between two measurements by the mean of the two values. Sibling pairs are organized into four groups: KIR identical, HLA identical (A); KIR identical, HLA disparate (B); KIR disparate, HLA-identical (C); and KIR disparate, HLA disparate (D). E and F compare relative frequency of CD94:NKG2A and KIR expression by NK cells of the 104 individual donors. E, The frequencies of the four measured KIR-positive subpopulations were simply summed (as subpopulations overlap, the total frequency exceeds 100%). F, The actual percentage of NK cells expressing any KIR or KIR combination was estimated from the individual frequencies and knowledge of their random association. The formula (100 (100 percent EB6 NKs)(100 percent DX27 NKs)(100 percent DX9 NKs)(100 percent DX31 NKs)) calculates the percentage of NK cells expected to express one or more KIR. The linear regression trend lines show the inverse correlation. Reconstitution of NK cell receptor repertoire following stem cell transplantation Reconstitution of NK cell receptor expression was followed in 18 patients undergoing stem cell transplantation for treatment of CML or acute myelogenous leukemia. Twelve of these transplants involved HLA-matched sibling donors, five involved HLA-matched unrelated donors (MUD), and one an unrelated donor that was HLA-C disparate, but otherwise HLA matched. Consistent with random segregation of parental KIR and HLA haplotypes, 3 (25%) of the 12 (25%) sibling donor/recipient pairs were KIR identical. All of the HLA-matched unrelated donor/recipient pairs were KIR disparate. The three donor/patient pairs with identical KIR genotypes had similar KIR repertoires (Fig. 6A), exhibiting summed differences comparable to those of healthy KIR- and HLA-identical sibling pairs (Fig. 3A). Thus, the hematologic malignancy suffered by the patients in these pairs had little effect on their KIR expression, indicating that, in general, the KIR repertoires measured in patients before transplant will reflect their healthy repertoires before the onset of malignant disease. The KIR-disparate donor/recipient pairs gave a wide range of summed differences, as did the MUD pairs (Fig. 6A), similar to those observed for healthy HLA-identical, KIR-disparate sibling pairs (Fig. 3C). Following transplantation, patients peripheral blood was sampled at intervals over the course of 1 year. Posttransplant KIR repertoires were compared with those of the donor and patient before transplant. The patients could be divided into three groups based on reconstitution of KIR and CD94:NKG2A expression. For the eight patients of group 1 (five MUD, three sibling donors), the reconstituted repertoire of KIR expression was like that of the donor and distinct from that of the patient before transplant (Fig. 6, B and C). Upon initial engraftment, few peripheral blood NK cells expressed KIR, while the majority were CD94:NKG2A positive. The frequency of KIR expression gradually increased during the subsequent 6 9 mo, reaching levels comparable to those of the donor. Concomitant with increasing KIR expression, the proportion of NK cells expressing CD94:NKG2A gradually lessened. Patients in this group (group 1) had good recovery from transplantation with no major clinical complications. Reconstitution in the five patients of group 2 (all sibling donors) was characterized by reduced frequencies of KIR-positive NK cells that persisted throughout the first year posttransplant, while CD94:NKG2A expression frequencies remained high (Fig. 6D). Although fewer NK cells expressed KIR, the hierarchy of KIR expression within the KIR-positive population resembled that of the donor. All but one of the group 2 patients experienced no major clinical complications during the first year posttransplant. A similar pattern of reconstitution was observed in patients receiving autologous stem cell transplants (Fig. 6E). Five patients (group 3) exhibited idiosyncratic patterns of NK cell receptor reconstitution (data not shown). They all suffered serious complications within the first year after transplant, including chronic GVHD, grade IV acute GVHD, and CML relapse. Discussion We developed a simple, robust, and quantitative method for comparing NK cell repertoires of KIR and CD94:NKG2A expression. Study of five healthy donors over a 1-year period showed that each person s repertoire was stable and unaffected by infection or other environmental stress. This raises our previous observation of KIR3DL1 expression stability (24) to a general principal. The validity of this conclusion of individual repertoire stability is underscored by the fact that the same method of comparison showed that each donor has a different repertoire. Thus, each donor s repertoire is genetically predetermined, a characteristic of innate defense mechanisms. Moreover, the variation from one donor to another shows that no one repertoire is optimal. From a practical standpoint, repertoire stability meant that reliable assessment of a person s repertoire could be based on a single blood donation. This made feasible the analysis of 100 individuals from whom correlations of NK cell repertoire difference with KIR and HLA genetics were made. By studying 85 sibling pairs from 36 different families, we were able to combine a fair sampling of the human population with valuable simplification of a complex genetics. Comparison of sibling pairs demonstrated that the KIR genotype is the dominant factor determining the repertoire

8 The Journal of Immunology 245 FIGURE 6. Distinct patterns of NK cell receptor repertoire reconstitution after stem cell transplantation. A, Comparisons between the pretransplant NK cell KIR repertoires of 18 allogeneic transplant patients and their MUD or sibling donors, identified as KIR-identical, HLA-identical sibling donor/patient pairs (f), KIR-disparate, HLA-identical sibling donor/patient pairs ( ), KIR-disparate, HLA-identical MUD/patient pairs ( ), and KIRdisparate, and HLA-disparate MUD/patient pairs (E). The HLA-disparate patient is mismatched for one HLA-C allele, but shares the same KIR epitopes as the donor. The same pretransplant NK cell KIR repertoires were compared with the 1 year posttransplant repertoires for the patients in group 1 (B); each symbol represents a different patient. Empty symbols show comparisons between posttransplant repertoire and recipient pretransplant repertoire; filled symbols represent comparisons between posttransplant repertoires and the donor repertoires. These results indicate that posttransplant KIR repertoires, in these patients, become like those of the donors. Posttransplant KIR and CD94:NKG2A NK cell expression frequencies of representative patients in group 1 (C) and group 2 (D) are compared with donor and recipient pretransplant expression; group 1 patients are characterized by a return to donor-like KIR expression, whereas for group 2, KIR expression remains depressed. E, Posttransplant KIR expression in a representative autologous transplant patient. Each receptorspecific Ab is shown with a different symbol and are EB6 ( ; KIR2DL1/ S1), DX27 (f; KIR2DL2 3/S2), DX9 ( ; KIR3DL1), DX31 (Œ; KIR3DL2), and Z199 (E; CD94: NKG2A). of KIR expression on NK cells, as is vividly seen from comparison of KIR-identical pairs with KIR-disparate pairs in the absence of any knowledge of HLA class I type (Fig. 3, E and F). Now evident is that the KIR genes themselves are the undefined, non-hla genes previously implicated in controlling KIR expression (24). In healthy individuals, the influence of the HLA genotype on NK cell KIR expression is subordinate to that of KIR, only being detectable when analysis was restricted to the subset of KIR-identical sibling pairs. Then it became clear that the KIR repertoires of KIR-identical, HLA-identical siblings were more similar than those of KIR-identical, HLA-disparate siblings. Importantly, the impact of HLA is to change the frequencies of KIR-expressing cells, not the surface levels of KIR expression. This provides good evidence that the HLA class I genotype imposes selection during development of the NK cell receptor repertoire and is consistent with functional observations showing that human NK cells express an inhibitory receptor for autologous HLA class I, though not necessarily for allogeneic HLA class I (4). Formally, it is possible that the HLA effect we observed is not due to HLA class I but to other linked genes of the HLA region. However, we consider this an unlikely possibility, given the well-established role of HLA class I polymorphisms in human NK cell receptor biology (23). The subtlety of HLA class I selection on NK cell repertoires arises because NK cells express multiple KIR, so that selection for cells expressing inhibitory receptors specific for autologous HLA class I causes only small reductions in the relative frequencies of cells expressing other KIR. This subtlety may have contributed to the failure of previous studies to see any influence of HLA class I on the expression of HLA-B- or HLA-C-specific KIR (24, 25). In those studies, comparison of the effect of the HLA class I difference was not made in the context of KIR identity and any effect due to HLA would have been obscured by the greater effects of the KIR genotype difference. Also critical was that the earlier studies used KIR epitope motifs of HLA-B and -C molecules as measures of HLA identity; these are simplified measures of KIR ligands which, in this study, did not reveal the HLA effect in selection of the NK cell KIR repertoire. Thus, the results obtained here are compatible with and resolve the seemingly paradoxical results obtained in previous investigations. Moreover, our data suggest that the HLA effect on the KIR repertoire may be governed by complex interactions between KIR and HLA molecules; these could include allelic fine specificities for the human KIR and as yet undiscovered KIR specificities. Clinical stem cell transplantation provides a system for examining the reconstitution of NK cell repertoires under conditions of HLA and/or KIR genetic difference. The patients we studied formed three groups, with those experiencing no major clinical complications following transplantation making up the first two. In the majority of allogeneic HLA-matched transplants, the patterns of KIR expression became like that of the donor, confirming the importance of the KIR genotype revealed in the comparison of healthy sibling pairs. In some allogeneic transplants and autologous transplants, the relative expression of the different KIR genes was like the donor, but the overall percentage of NK cells expressing KIR was reduced compared with the donor. One possibility was that CD56 dim KIR cells were replaced by CD56 bright KIR cells (45, 46). This did not seem to be the case, because lack of KIR expression was seen in both the CD56 bright and CD56 dim populations. In contrast, allogeneic transplants between HLA-matched unrelated individuals tended to reconstitute NK cell populations with frequencies of KIR expression comparable to those of the donor. Thus, in the transplant situation, some degree of genetic incompatibility may facilitate induction of KIR expression. A potentially related phenomenon is that in vitro culture of human NK

9 246 GENETIC CONTROL OF HUMAN NK CELL REPERTOIRE cell precursors with xenogeneic mouse feeder cells facilitated induction of KIR expression (47). An alternative explanation for the distinct reconstitution patterns was the more common use of fractionated total-body irradiation in the conditioning regimens of group 1 patients (seven of eight) than group 2 patients (one of five), which may have more thoroughly ablated recipient hemopoietic cells and/or had a more damaging effect on the stromal environment The third group of patients was those suffering major clinical complications following transplantation. Their idiosyncratic patterns of NK cell receptor expression suggest that GVHD, early relapse, and other complications or clinical interventions can influence NK cell receptor expression. This is consistent with a previous study showing lower frequencies of KIR expression in patients with chronic GVHD compared with those without, although, in that study, the patient and donor KIR types were not taken into account (48). Whereas HLA class I modifies the frequency of KIR-expressing cells in NK cell repertoires, shown here in healthy donors, it does not affect expression levels. This property of KIR contrasts with observations made of the Ly49 molecules of mice. Surface levels of Ly49 are lower in the presence of cognate MHC class I ligands than in their absence (49 51). The proportions of NK cells expressing various Ly49 molecules are also affected by the class I environment; the expression of a self-specific Ly49 decreases the subsequent expression of additional Ly49 receptors capable of binding autologous MHC. The results support a scheme whereby each NK cell expresses receptors stochastically until a threshold level of functional interaction with a self-mhc molecule is reached (51 55). Whether the observed differences in humans and mice reflect their biology or are due to the very different, and often complementary, approaches taken to their investigation remains to be seen. KIR are expressed in stochastic fashion, which does not always end with expression of an inhibitory KIR with specificity for autologous HLA class I. For a proportion of NK cell clones, the CD94:NKG2A receptor fills that role. Comparison here of 104 healthy donors revealed a statistically significant inverse correlation between the proportions of NK cells expressing KIR and CD94:NKG2A. This relationship was also apparent during poststem cell transplant recovery of NK cell populations. Together, these data provide direct evidence for the two types of receptors sharing common purpose and for some coordination in their differential expression during NK cell development. Long-term follow-up of patients having NK cell receptor repertoires with diverse KIR specificities (group 1) vs repertoires biased toward CD94: NKG2A usage (group 2) may provide an indication of the relative importance of KIR and CD94:NKG2A to NK cell function in vivo. That the repertoire of NK cell CD94:NKG2A expression is stable in healthy individuals but varies between them is consistent with genetic control, although data from both healthy sibling pairs and stem cell transplant patients suggests that regulation of CD94: NKG2A expression may be more complex than that of KIR. Variability in the CD94 and NKG2A genes could contribute to this effect. The evidence available shows these genes to be highly conserved in the human population, although the effect of potential allelic differences outside the coding region on CD94:NKG2A expression cannot be completely dismissed (14, 56, 57). Our study suggests that patterns of CD94:NKG2A expression are influenced by the combination of HLA and KIR genotypes, with the former having greater effect. Again, this is consistent with CD94:NKG2A and KIR sharing the function of providing inhibitory receptors for NK cells and for there being coordination in expression of the two types of genes. One explanation for the greater influence of HLA than KIR genotype on CD94:NKG2A expression is that HLA genotypes vary in the number of KIR ligands they provide, whereas the great majority of KIR genotypes provide all of the different types of inhibitory KIR with specificity for HLA class I. Consequently, the HLA class I genotype will usually dictate the number of KIR that can serve as inhibitory receptors for autologous HLA class I and thus the proportion of NK cells needing CD94:NKG2A expression to be tolerant of self. A second, and not necessarily mutually exclusive, possibility is that the HLA effect is due to polymorphism affecting either the expression or function of HLA-E, the ligand for CD94:NKG2A. Acknowledgments We thank Dr. Grumet and the Stanford Medical Center Histocompatibility Laboratory staff; Dr. Rodriguez and the City of Hope Bone Marrow Transplant Program; Dr. Khakoo for help in collecting samples from healthy individuals; Drs. Shizuru, Blume, and Negrin and Kathryn Tierney for collection of transplant patient samples and helpful advice; Dr. Uhrberg for providing KIR epitope-typing primers; and Dr. Brown for help and advice with statistical analysis. References 1. Trinchieri, G Biology of natural killer cells. Adv. Immunol. 47: Kumar, V., T. George, Y. Y. Yu, J. Liu, and M. Bennett Role of murine NK cells and their receptors in hybrid resistance. Curr. Opin. Immunol. 9: Ruggeri, L., M. Capanni, M. Casucci, I. Volpi, A. Tosti, K. Perruccio, E. Urbani, R. S. Negrin, M. F. Martelli, and A. Velardi Role of natural killer cell alloreactivity in HLA-mismatched hematopoietic stem cell transplantation. Blood 94: Valiante, N. M., M. Uhrberg, H. G. Shilling, K. Lienert-Weidenbach, K. L. Arnett, A. D Andrea, J. H. Phillips, L. L. Lanier, and P. Parham Functionally and structurally distinct NK cell receptor repertoires in the peripheral blood of two human donors. Immunity 7: Lanier, L. L NK cell receptors. Annu. Rev. Immunol. 16: Raulet, D. H., W. Held, I. Correa, J. R. Dorfman, M. F. Wu, and L. Corral Specificity, tolerance and developmental regulation of natural killer cells defined by expression of class I-specific Ly49 receptors. Immunol. Rev. 155: Uhrberg, M., N. M. Valiante, B. P. Shum, H. G. Shilling, K. Lienert-Weidenbach, B. Corliss, D. Tyan, L. L. Lanier, and P. Parham Human diversity in killer cell inhibitory receptor genes. Immunity 7: Suto, Y., Y. Ishikawa, M. Kasahara, F. Kasai, T. Yabe, T. Akaza, and T. Juji Gene arrangement of the killer cell inhibitory receptor family on human chromosome 19q13.4 detected by fiber-fish. Immunogenetics 48: Wende, H., M. Colonna, A. Ziegler, and A. Volz Organization of the leukocyte receptor cluster (LRC) on human chromosome 19q13.4. Mamm. Genome 10: Wilson, M. J., M. Torkar, A. Haude, S. Milne, T. Jones, D. Sheer, S. Beck, and J. Trowsdale Plasticity in the organization and sequences of human KIR/ ILT gene families. Proc. Natl. Acad. Sci. USA 97: Borrego, F., M. Ulbrecht, E. H. Weiss, J. E. Coligan, and A. G. Brooks Recognition of human histocompatibility leukocyte antigen (HLA)-E complexed with HLA class I signal sequence-derived peptides by CD94/NKG2 confers protection from natural killer cell-mediated lysis. J. Exp. Med. 187: Braud, V. M., D. S. Allan, C. A. O Callaghan, K. Soderstrom, A. D Andrea, G. S. Ogg, S. Lazetic, N. T. Young, J. I. Bell, J. H. Phillips, et al HLA-E binds to natural killer cell receptors CD94/NKG2A, B and C. Nature 391: Brooks, A. G., F. Borrego, P. E. Posch, A. Patamawenu, C. J. Scorzelli, M. Ulbrecht, E. H. Weiss, and J. E. Coligan Specific recognition of HLA-E, but not classical, HLA class I molecules by soluble CD94/NKG2A and NK cells. J. Immunol. 162: Shum, B. P., L. R. Flodin, D. G. Muir, R. Rajalingam, S. I. Khakoo, S. Cleland, L. A. Guethlein, M. Uhrberg, and P. Parham Conservation and variation in human and common chimpanzee CD94 and NKG2 genes. J. Immunol. 168: Vance, R. E., D. M. Tanamachi, T. Hanke, and D. H. Raulet Cloning of a mouse homolog of CD94 extends the family of C-type lectins on murine natural killer cells. Eur. J. Immunol. 27: Ho, E. L., J. W. Heusel, M. G. Brown, K. Matsumoto, A. A. Scalzo, and W. M. Yokoyama Murine Nkg2d and Cd94 are clustered within the natural killer complex and are expressed independently in natural killer cells. Proc. Natl. Acad. Sci. USA 95: Lohwasser, S., P. Hande, D. L. Mager, and F. Takei Cloning of murine NKG2A, B and C: second family of C-type lectin receptors on murine NK cells. Eur. J. Immunol. 29: Yabe, T., C. McSherry, F. H. Bach, P. Fisch, R. P. Schall, P. M. Sondel, and J. P. Houchins A multigene family on human chromosome 12 encodes natural killer-cell lectins. Immunogenetics 37: Renedo, M., I. Arce, A. Rodriguez, M. Carretero, L. L. Lanier, M. Lopez-Botet, and E. Fernandez-Ruiz The human natural killer gene complex is located on chromosome 12p12 p13. Immunogenetics 46:307.