(KIR) and HLA Class I Genes

ORIGINAL ARTICLE Signatures of natural selection and coevolution between killer cell immunoglobulin-like receptors (KIR) and HLA class I genes

KJ Guinan1, RT Cunningham2,4, A Meenagh2, A Gonzalez2, MM Dring1, BW McGuinness1, D Middleton3 and CM Gardiner1 1School of Biochemistry and Immunology, Trinity College, Dublin, Ireland; 2Histocompatibility and Immunogenetics Laboratory, Belfast City Hospital, Belfast, Northern Ireland, UK and 3Royal Liverpool University Hospital and School of Infection and Host Defense, Liverpool University, Liverpool, UK

Natural killer (NK) cells are lymphocytes of the innate immune system. In humans, NK cell activities are partly controlled by the diverse killer immunoglobulin-like receptor (KIR) gene family. The importance of NK cells in both immunity to infection and reproduction makes KIR strong candidates for genes undergoing dynamic evolution in the human genome. Using high- resolution allelic typing, we investigated the potential role of natural selection in the diversification of KIR in the Irish population. Higher diversity than expected is observed at several loci, consistent with a history of balancing selection acting to maintain several allelic variants at high frequency in the population. KIR diversity is enhanced further at the haplotype level with functional polymorphisms at KIR2DL4, KIR3DL1 and KIR2DS4 defining nine ‘core’ haplotypes. Analysis of these core haplotypes in combination with human leukocyte antigen (HLA) class I ligands revealed several nonrandom associations. In particular, the KIR:HLA association for the core haplotype defined by KIR3DL1*01502 was female specific and a likely consequence of negative selection acting against KIR3DL1*01502 on an HLA-C1/C1 background. Many of the associations between KIR and HLA in the Irish differ from those previously reported, which argues against universal selective pressures for specific KIR:HLA combinations in diverse human populations. Genes and Immunity (2010) 11, 467–478; doi:10.1038/gene.2010.9; published online 4 March 2010

Keywords: NK cells; killer cell immunoglobulin-like receptors; evolution; KIR haplotypes; HLA; natural selection

Introduction infections. This system is also considered important during pregnancy when NK cells interact with fetal and To ensure the survival of a species, individuals within trophoblast cells expressing HLA class I to promote populations must possess genetic characteristics, which vascular remodeling and placentation.6–9 KIR are highly enable them to survive infectious disease, reproduce polymorphic and individuals vary in the type and and pass on their genes.1 Natural killer (NK) cells have number of KIR genes they inherit.10 Two broad classifica- been shown to be key immune effector cells in the fight tions of KIR haplotypes exist: ‘A’ haplotypes are against pathogen but they have also been shown to have restricted in terms of the number of genes they possess important roles during pregnancy.2,3 These contrasting and primarily encode for inhibitory receptors (KIR2DL1, activities involve the killer cell immunoglobulin-like KIR2DL3, KIR3DL1, KIR3DL2, KIR2DL4 and the activat- receptors (KIR) and human leukocyte antigen (HLA) ing gene, KIR2DS4). In contrast, ‘B’ haplotypes have a receptor:ligand system. KIR are a family of receptors higher gene content (ranges from 7–16 genes) and encoded on human chromosome 19 that are expressed on although they have genes encoding for inhibitory the surface of NK cells.4,5 They enable NK cells to receptors (including the characteristic KIR2DL5 gene), recognize and respond to HLA class I downregulation on they also have additional activating genes including target cells, an important hallmark of many viral KIR2DS1, KIR2DS2, KIR2DS3, KIR2DS5 and KIR3DS1.11,12 NK cell function is highly dependent on the type of KIR and HLA that individuals inherit, and a range of Correspondence: Dr CM Gardiner, School of Biochemistry and compound genotypes conferring different NK cell Immunology, Trinity College, Dublin 2, Ireland. responsiveness phenotypes have been defined.13–15 The E-mail: [email protected] dual role of KIR:HLA genes in immunity to infection and 4 Current address: Institute of Agri-Food & Land Use, Queen’s reproduction makes them good candidates for genes University, David Keir Building, Stranmillis Road, Belfast BT9 5AG, Northern Ireland. undergoing natural selection and coevolution in the Received 28 September 2009; revised 2 December 2009; accepted 6 human genome. Characteristics that promote survival January 2010; published online 4 March 2010 against pathogen and successful reproduction are likely KIR and HLA genes are coevolving KJ Guinan et al 468 to be increased in frequency in humans, whereas genes that hinder these processes may be reduced in frequency. Analysis is complex as KIR and HLA genes are inherited on different chromosomes, known interactions are epistatic and not all functional interactions have been defined. However, there is evidence for coevolution between these genes as a positive association between the presence of inhibitory KIR and their HLA ligands, and a strong negative association between activating KIR and their HLA ligands is observed globally across human populations.16 Furthermore, we have previously found evidence that KIR and HLA class I genes are coevolving within Europe and have shown a balance between inhibitory and activating KIR repertoires, and their ligands, in different regions of Europe.17 Supporting the view that these genes are coevolving, a gender bias in terms of nonrandom associations between KIR core haplotypes and HLA class I has been found in the Japanese population.18 Of particular interest was the finding that several associations between KIR and HLA genes were limited to females supporting the view that reproduction is a strong selective pressure acting on KIR genes.1,7,18 It is not known whether the particular combinations of KIR and HLA found in Japanese women reflects a universal mechanism to promote reproductive success and/or resistance to pathogen within the human species or whether it is a consequence of more local Figure 1 Balancing selection drives high KIR diversity in the Irish environmental and genetic selective pressures. To test the population. (a) Donors were typed to the allele level for five hypothesis that there is a universal selective pressure polymorphic KIR (sample numbers indicated separately for each locus) and the frequencies (F) of each are shown. Ambiguous allele operating to promote particular KIR and HLA combina- typing is indicated. (b) Diversity at these five polymorphic KIR gene tions in humans, potentially in a gender-specific manner, loci was assessed using the Ewens Watterson test. The black line we performed high-resolution allelic analysis of KIR and indicates heterozygosity expected under evolutionary conditions of HLA class I genes in the Irish population and investi- neutrality. Values above the black line indicate deviation from gated their coevolution with respect to known functional neutrality consistent with balancing selection, whereas deviation interactions. below the line reflects processes of purifying selection. Significant deviation from neutrality is indicated by *P-value o0.05, **P-value o0.01, ***P-value o0.0001. Results To investigate the role of natural selection acting on KIR indicated by blue-rimmed boxes (Po0.05). A striking genes in a genetically homogenous population, Irish relationship is seen when we look at LD between alleles donors (n ¼ 240) were typed to the allele level for of KIR2DL4 and KIR3DL1. For a given allele of KIR3DL1 five polymorphic loci (KIR2DL3, KIR2DL4, KIR3DL1, (with the exception of KIR3DL1*007/*020 that is KIR2DS4 and KIR3DL2 loci). Diversity at these loci was found with two alleles of KIR2DL4), there is only ever assessed using the BOTTLENECK program,19 which positive LD with a single allele at the KIR2DL4 locus, for compares heterozygosity at an individual locus to that example, KIR3DL1*00101 is only found in strong positive expected under neutral conditions. Each KIR locus LD with KIR2DL4*00801. Alleles of KIR2DL4 are exhibits heterozygosity greater than expected and, in found in positive LD with a restricted and generally particular, this is statistically significant for KIR2DL4 nonoverlapping subset of KIR3DL1 alleles, for example, (Po0.0001), KIR3DL1 (Po0.01) and KIR2DS4 (Po0.05) KIR2DL4*00801 is found in strong LD with both loci (Figure 1). A higher level of heterozygosity than KIR3DL1*00101 and KIR3DL1*009 and neither of these expected is consistent with a history of balancing KIR3DL1 alleles are found in positive LD with any other selection acting to maintain several alleles at high KIR2DL4 allele. Similarly, alleles of KIR2DS4 are found frequency in the population. No evidence of purifying in strong LD with alleles of KIR2DL4 giving unique LD selection on any single allele was detected. associations between these genes. In combination with We next analyzed the cohort for evidence of linkage KIR2DL4/KIR3DL1 data, this predicts a three-gene core disequilibrium (LD) between alleles of KIR genes using haplotype defined by alleles of KIR2DL4–KIR3DL1– the recently developed MIDAS program.20 This involved KIR2DS4. Indeed, LD analysis of individual KIR3DL1 a systematic pairwise comparison of the frequency of alleles shows that they are almost uniquely associated each unique combination of alleles of the KIR2DL3, with a single KIR2DS4 subtype (although individual KIR2DL4, KIR3DL1, KIR2DS4 and KIR3DL2 genes. The KIR2DS4 alleles may be associated with more than output is graphically represented in Figure 2. Positive LD one KIR3DL1 allele). Thus, LD was used to define between alleles is indicated by a red box with asterisk (D0 KIR2DL4–KIR3DL1–KIR2DS4 as a core haplotype 0.8–1.0, Po0.05), an empty red box (D0 0.6–0.8, Po0.05) region in the Irish as has previously been found in the or a pink box (D0 0.2–0.6, Po0.05), and negative LD is Japanese population.18 Although still positive, LD between

Genes and Immunity KIR and HLA genes are coevolving KJ Guinan et al 469

Figure 2 Linkage disequilibrium (LD) between KIR loci. LD was assessed between every combination of two alleles of five polymorphic KIR loci using the MIDAS software and the output generated is shown graphically. Positive LD between alleles is indicated by red boxes with asterisks (high positive LD with D0 range of 0.8–1.0), empty red boxes (positive LD with D0 range of 0.6–0.8) or pink boxes (reduced positive LD with D0 range of 0.2–0.6). Negative LD is indicated by blue-rimmed boxes (D0 range of À1.0 to À0.6). An example of how to read this figure is shown in the top left-hand square for alleles KIR2DL3*005 and KIR2DL4*00801, which are in weak LD. Significance was determined by w2-test using a Yates correction. All positive and negative values shown are statistically significant (P-values o0.05). The order in which the genes appear on the chromosome is also shown.

KIR3DL2 and the core haplotype genes breaks down that KIR2DL3 is a source of diversity in the Irish somewhat and alleles of KIR3DL2 are generally found population. associated with several alleles of an individual locus, for Using KIR gene and allele profiles of each donor, we example, KIR3DL2*002 is found with KIR3DL1*002 and used PHASE 2.121,22 to reconstruct KIR haplotypes and KIR3DL1*01502. In some cases, alleles of KIR3DL2 can be the 25 most frequent of these are shown in Table 1. used to define extended haplotypes, for example, Haplotypes that were identified in one or more donors KIR3DS1 is always found with KIR3DL2*007. There is were included in further analysis. These haplotypes little positive LD with KIR2DL3 and genes of the core occurred at a combined frequency of 85.4% in haplotype (absence of red boxes in Figure 2), indicating our population. The most common haplotype, a ‘B’

Genes and Immunity KIR and HLA genes are coevolving KJ Guinan et al 470 Table 1 KIR haplotypes in the Irish population

Haplotype KIR KIR KIR KIR KIR KIR KIR KIR KIR KIR KIR KIR KIR KIR KIR Population 3DL3 2DS2 2DL2/3 2DL5b 2DL1 2DP1 3DP1 2DL4 3DL1 2DL5a 2DS3 2DS5 2DS1 2DS4 3DL2 Frequency 1 B 001 005 3DS1 001 002 002 007 0.063 2 A 001 00102a 01502 00101 002 0.061 3 A 002/7 00801 00101 003 001/013 0.053 4 A 001 011 005 007 001/013 0.041 5 A 001 00802 004 006 003 0.034 6 A 001 00801 00101 003 001/013 0.029 7 A 001 00103a 008 003 009 0.022 8 A 001 00102a 002 00101 002 0.021 9 B 002/7 005 3DS1 001 002 002 007 0.019 10 B 002/7 005 3DS1 001 002 002 00102 007 0.018 11 A 002/7 00802 004 006 005 0.018 12 B 2DL2 00102a 002 00101 002 0.017 13 B 2DL2 002 00102a 01502 00101 002 0.017 14 B 2DL2 002 005 3DS1 005 002 007 0.017 15 A 002/7 00102a 002 00101 002 0.016 16 B 2DL2 00801 00101 003 001/013 0.015 17 B 2DL2 002 00102a 002 00101 002 0.015 18 A 002/7 006 007b 004 008 0.015 19 A 002/7 011 005 007 001/013 0.013 20 B 2DL2 00801 00101 003 011 0.013 21 A 002/7 00802 004 006 009 0.012 22 A 001 00802 004 006 005 0.012 23 A 002/7 00802 004 006 005 0.011 24 B 2DL2 00102a 002 00101 002 0.011 25 B 2DL2 005 3DS1 001 002 002 007 0.011

Abbreviation: KIR, killer immunoglobulin-like receptor. Allele and genotype data were used to reconstruct haplotypes using Phase 2.1 and the 25 most common haplotypes are shown (accounting for 57.5% of the total). The presence of a KIR gene is indicated by a gray-filled box, whereas a white box indicates gene absence. Genes are ordered to reflect their chromosomal location. The individual alleles present at several loci are indicated. A or B haplotype status is indicated in the second column and the frequencies of the predicted haplotypes are indicated in the rightmost column. aAllele type inferred from KIR2DL4*00102/*00103 ambiguity based on segregation analysis of a related family cohort. bAllele type inferred from KIR3DL1*007/020 ambiguity using segregation analysis of a related family cohort.

haplotype characterized by the presence of KIR2DL3*001, 6.3% of the Irish population. This relatively low KIR2DL4*005, KIR3DS1, KIR3DL2*007 and the absence frequency indicates that there is significant haplotype of KIR2DS4 (haplotype 1 in Table 1), was present in only diversity in the Irish population as previously reported

Genes and Immunity KIR and HLA genes are coevolving KJ Guinan et al 471

Figure 3 Haplotype homozygosity analysis identifies signatures of natural selection operating on KIR genes. (a) Unique core haplotypes identified by KIR2DL4–KIR3DL1–KIR2DS4 polymorphism and their frequencies are shown for the Irish population. ---Indicates the absence of KIR2DS4 gene. (b) Haplotype homozygosity for the nine high-frequency core haplotypes defined in (a) was examined and graphed using the same symbols as in (a). Core haplotypes (KIR2DL4–KIR3DL1–KIR2DS4) are represented in the central region by big black square blocks. Haplotype homozygosity is shown on the y axis for centromeric (KIR2DL3) and telomeric (KIR3DL2) ends of the core haplotypes with each symbol representing a unique core haplotype as defined in (a). (c, d) The frequencies of core haplotypes are plotted against haplotype homozygosity at centromeric (c) and telomeric (d) ends. Each symbol represents a unique core haplotype as defined in (a). The core haplotype defined by KIR3DL1*01502 is highlighted using an arrow in (c). Analysis is based on haplotypes present in one or more donors (representative of 85.4% of the total population).

in a study of Northern Irish families.23 This contrasts the cell surface. Overall, the allele present at the with the Japanese in which their most common haplo- KIR3DL1 locus is an excellent marker for use in type (our second most common haplotype at 6.1%) was haplotype tagging and can define individual, high- present in 35.7% of the population.18 We defined the frequency haplotype cores and the alleles they contain common core haplotypes using the allele present at each at KIR2DL4 and KIR2DS4 loci (Figure 3a). of the loci, KIR2DL4, KIR3DL1 and KIR2DS4, and those Defining KIR2DL4–KIR3DL1–KIR2DS4 as our haplo- present at 42.5% in our population are shown in type core region, the data were tested for haplotype Figure 3a. Six of the core haplotypes are found at a homozygosity, that is, allelic homogeneity, centromeric frequency of between 10 and 20% (as shown in and telomeric of the core region. This analysis takes each Figure 3a), suggesting that balancing selection is acting of the core allele haplotypes present in the Irish to maintain these different cores at relatively high levels population (defined in Figure 3a) and examines them in the population. Five of the six core haplotypes defined for the extent of variation in genes either telomeric in the Japanese were also found in the Irish, but the Irish (KIR3DL2) or centromeric (KIR2DL3) of the core. For a had additional novel core haplotypes including those given high-frequency core haplotype, high homozygos- defined by KIR3DL1*004 and KIR3DL1*002. The ity (that is, limited allelic variation) at loci outside both KIR3DL1*004 core haplotype is interesting functionally ends of the core is indicative of positive selection acting as all three allelic variants of KIR2DL4–KIR3DL1– on a haplotype or its components.24 Supporting our LD KIR2DS4 encode for proteins that are not expressed at data, five of the nine core haplotypes exhibit high

Genes and Immunity KIR and HLA genes are coevolving KJ Guinan et al 472 diversity (low homozygosity) centromeric of the core at promote immunity to pathogen and successful reproduc- KIR2DL3 (Figure 3b), and another three exhibit inter- tion.1,7,16,18,28 This type of analysis is complicated as both mediate levels of diversity at KIR2DL3, confirming it as a gene families are polygenic and highly polymorphic, and locus contributing to haplotype diversity in the Irish they segregate independently on different chromosomes. population. Diversity telomeric of the KIR2DL4– All our donors were typed for HLA class I and stratified KIR3DL1–KIR2DS4 core is more restricted. Only one of according to the known ligands provided to KIR the nine core haplotypes (defined by KIR3DL1*004) is receptors. In brief, HLA-C1 refers to HLA-C alleles that characterized by high diversity at the telomeric KIR3DL2 provide a ligand for the KIR2DL3 KIR receptor (includ- locus and a further three have intermediate levels of ing its common allele, KIR2DL2); HLA-C2 refers to HLA- diversity. In contrast to KIR2DL3, five of the core C alleles that provide a ligand for the KIR2DL1 receptor; haplotypes have high homozygosity with respect to and HLA-Bw4 refers to HLA-B alleles that provide a KIR3DL2 alleles, that is, are often found with a given ligand for KIR3DL1 receptor. As an approach to allele of KIR3DL2. Thus, despite being more poly- investigate coevolution of these genes, we tested the morphic than KIR2DL3, KIR3DL2 is a gene locus that potential for nonrandom associations between KIR and restricts diversity of core haplotypes in the Irish. HLA class I genes in the Irish. Using the allele present at Under conditions of neutrality, alleles present at high the KIR3DL1 locus to identify individual core haplotypes frequency in populations are expected to be present on a (as defined in Figure 3a), we found that three high wide range of haplotypes because of increased numbers frequency, functionally distinct core haplotypes associate of recombination events occurring over time. Conversely, with HLA in a nonrandom manner. Examination of the under conditions of natural selection, an allele selected KIR3DL1*01502 core showed it was present at an for and driven to high frequency is expected to be increased frequency on an HLA-C2 homozygous back- present on a more restricted range of haplotypes.24 We ground (Table 2 and summarized in Table 3a). As HLA examined extended haplotypes, centromeric and telo- alleles defining HLA-C1 and HLA-C2 NK cell ligands meric of the core for evidence of natural selection. When are mutually exclusive, an increased frequency of a we graph the frequency of core haplotypes versus particular KIR on an HLA-C2 background is associated haplotype homozygosity at the centromeric KIR2DL3 with a reduced frequency on an HLA-C1 background. locus, we see no evidence of selection centromeric of the Another high-frequency core, defined by the core haplotype (Figure 3c). The exception to this is the KIR3DL1*00101 allele, was associated with HLA-C1 core haplotype defined by KIR3DL1*01502 that although homozygosity (Tables 2 and 3a). Finally, when examined present at a high frequency (10.6%) in the Irish on an HLA-C2-negative background (to account for LD population is found predominantly with the between HLA-Bw4 and HLA-C2), KIR3DS1 was nega- KIR2DL3*001 allele. In contrast, the pattern obtained is tively associated with the presence of its putative ligand, quite different telomeric of the core. Although the low- HLA-Bw4 (Tables 2 and 3a). Nonrandom associations of frequency core haplotypes show high homozygosity as KIR and HLA suggest that these gene families are expected, core haplotypes at high frequency in the Irish coevolving. population have higher than expected homozygosity at We also examined alleles at other KIR loci and found the telomeric KIR3DL2 locus despite the fact that increased frequencies of KIR2DL4*00102, KIR2DS4*00101 KIR3DL2 is more polymorphic than KIR2DL3 and KIR2DL3*001 alleles on an HLA-C2 homozygous (Figure 3d). Although lower recombination rates telo- background, similar to that seen for KIR3DL1*01502 meric of the core may give rise to the trend observed,25,26 (Tables 2 and 3b). These four alleles are in high LD the data may alternatively suggest positive selection together and are present on the second most frequent acting on these cores. In support of this, the high- haplotype (defined by KIR3DL1*01502) in the popula- frequency core haplotype defined by KIR3DS1 is only tion. This is also the haplotype that is characterized by ever found with KIR3DL2*007 and this combination of high homozygosity of polymorphic genes both centro- KIR3DL1 and KIR3DL2 alleles has previously been meric and telomeric of the core haplotype. Together, this found together in the Mexican and Japanese popula- suggests that selection may be operating on a particular tions.18,27 The maintenance of these combinations of component of this core haplotype or the entire haplo- alleles in such divergent populations is more likely to type. In a similar manner, KIR2DS1, KIR2DS5 and reflect positive selection. Strikingly, the core haplotype KIR2DL5 genes, which are all found in high LD with defined by KIR3DL1*01502 has restricted diversity at KIR3DS1, had similar nonrandom associations with HLA KIR3DL2 and is uniquely associated with the to that seen for KIR3DS1 (data not shown). Thus, it seems KIR3DL2*002 allele. Thus, the KIR3DL1*01502 core that there are HLA selection pressures acting on KIR haplotype has limited diversity both centromeric genes but our data do not distinguish if it is acting to (KIR2DL3) and telomeric (KIR3DL2) of the core. This is select a full haplotype or a component thereof. also evidenced in Figures 3c and d, in which, However, if we stratify these data based on gender, we KIR3DL1*01502 is the only high-frequency core haplo- find that the core haplotype defined by KIR3DL1*01502 type to have high homozygosity at both ends of the core (and its individual component alleles, KIR2DL4*00102 haplotype, suggesting that there is possible positive and KIR2DS4*00101), is increased on an HLA-C2 homo- natural selection acting on this KIR haplotype. These zygous background and that this is gender specific and findings at the haplotype level are supported by our LD increased in females (Tables 2 and 3b). Differences were analysis at the allelic level and together provide strong confirmed by directly comparing the frequencies of evidence that natural selection is acting to shape KIR KIR:HLA compound genotypes in males and females, diversity in the Irish population. in which the corresponding negative association with Evidence is emerging to support the tenet that KIR and HLA-C1 was found, that is, KIR2DL4*00102 and HLA class I genes are coevolving in directions that KIR2DS4*00101 frequencies were significantly reduced

Genes and Immunity KIR and HLA genes are coevolving KJ Guinan et al 473 Table 2 Nonrandom associations of KIR and HLA class I genes

KIR core haplotype Group KIR event HLA-C counts P-value w2/FE HLA-B counts P-value w2/FE or gene C2++ C1++ Bw4+ Bw4À

KIR2DL4*00102 3DL1*01502+ 10 18 0.0195* 5.452 26 14 0.3629 0.8278 Total 3DL1*01502À 17 90 111 83 KIR3DL1*01502 0.0011 10.58 Female 3DL1*01502+ 7 7 ** 16 5 0.0807 3.052 3DL1*01502À 542 5141 KIR2DS4*00101 Male 3DL1*01502+ 3 11 1 FE 10 9 0.6305 0.2314 3DL1*01502À 12 47 58 41

KIR2DL4*00801 3DL1*00101+ 4 43 0.0224* FE 20 23 0.1863a 1.747 Total 3DL1*00101À 23 65 22 43 a KIR3DL1*00101 Female 3DL1*00101+ 2 17 0.2268 1.461 7 10 0.9702 0.0014 3DL1*00101À 10 32 13 19 0.0360 FE 0.0488 ,a 3.882 KIR2DS4*003 Male 3DL1*00101+ 2 26 * 13 13 * 3DL1*00101À 13 32 8 24

KIR2DL4*005 3DS1+ 9 47 0.3736 0.7916 11 36 0.0059**,a 7.592 Total 3DS1À 18 63 31 32 KIR3DS1 0.0106 ,a 6.53 Female 3DS1+ 3 23 0.207 FE 5 18 * 3DS1À 926 1511 a KIR2DS4 negative Male 3DS1+ 6 23 0.9426 0.0052 5 18 0.0756 3.156 3DS1À 936 1620

0.0380 FE Total 2DL3*001+ 22 68 * 84 61 0.7299 0.1192 2DL3*001À 440 5335 KIR2DL3*001 Female 2DL3*001+ 9 30 0.3024 FE 40 29 0.6129 0.256 2DL3*001À 218 2716 2DL3*001+ 13 37 0.1226 FE 44 30 0.7 0.1485 Male 2DL3*001À 221 2419

0.0053 7.771 Total 2DL4*00102+ 19 45 ** 67 40 0.2333 1.421 2DL4*00102À 866 7259 KIR2DL4*00102 0.0020 FE Female 2DL4*00102+ 10 15 ** 30 14 0.1722 1.864 2DL4*00102À 234 3730 2DL4*00102+ 9 29 0.4189 0.6534 36 26 0.7944 0.0679 Male 2DL4*00102À 631 3427

0.0295 4.738 Total 2DS4*00101/2+ 14 38 0.0796 3.073 61 29 * 2DS4*00101/2À 12 70 78 68 KIR2DS4*00101/2 0.0264 4.929 0.0262 4.941 Female 2DS4*00101/2+ 7 12 * 28 9 * 2DS4*00101/2À 536 4135 2DS4*00101/2+ 7 25 0.6411 0.2172 32 20 0.4134 0.6124 Male 2DS4*00101/2À 333 3731

Abbreviations: FE, Fisher’s exact test; HLA, human leukocyte antigen; KIR, killer immunoglobulin-like receptor. Frequencies of KIR and HLA compound genotypes were tested for evidence of nonrandom association by w2 analysis or Fisher’s exact test. The top part of the table shows analysis for three core KIR haplotypes defined by KIR3DL1 polymorphism and the bottom part shows data for alleles present on the KIR haplotype defined by KIR3DL1*01502. Positive and negative counts for each comparison are shown. The P-values are indicated and those that are statistically significant are highlighted in bold. C2++ and C1++ refer to homozygosity for HLA-C2 and HLA-C1, respectively. Bw4+ or Bw4– refers to the presence or absence of HLA-Bw4. aIndicates analysis carried out on a HLA-C2 negative background. *Po0.05, **Po0.01. in females on an HLA-C1C1 background (Table 4). The an HLA-C1 homozygous background in females and that carrier frequency of the KIR3DL1*01502 allele between this selective pressure is operating outside the KIR3DL1 males and females was similar (data not shown). locus on other components of the haplotype, for However, although not reaching statistical significance, example, KIR2DL4 and/or KIR2DS4. two other components of this core haplotype are lower in females compared with males: KIR2DL4*00102 has a carrier frequency of 34.5% in females (n ¼ 40) and 50.0% Discussion in males (n ¼ 58), and KIR2DS4*00101 has a carrier frequency of 31.2% in females (n ¼ 34) and 43.3% in There is strong evidence emerging that balancing males (n ¼ 52). This observation, combined with a lack of selection is the main force shaping the evolution of evidence for nonrandom associations with HLA in genes involved in innate immunity.29 Balancing selection males, suggests that there may be negative selection increases the range and diversity of receptor systems against the presence of the KIR3DL1*01502 haplotype on available to the immune system to fulfil its wide-ranging

Genes and Immunity KIR and HLA genes are coevolving KJ Guinan et al 474 functions. Our analysis of the NK cell-associated KIR genes, which make up the KIR ‘core haplotype’. genes further supports this tenet. In the Irish population, Individually, these three loci are characterized by heterozygosity excess (that is, high diversity) is observed polymorphic diversity that affects receptor expression at each polymorphic KIR locus examined and this was and function. Three main lineages of the KIR3DL1 gene particularly evident for KIR2DL4, KIR3DL1 and KIR2DS4 (KIR3DL1*015-, KIR3DL1*005- and KIR3DS1-like) have been defined28 and shown to be maintained in humans by balancing selection. The Irish have strong representa- Table 3 Summary of nonrandom associations between KIR and tion of each of the three main KIR3DL1 lineages and also HLA class I genes. from a fourth lineage containing newer recombinant (a) KIR3DL1 alleles. One of these, KIR3DL1*00101, is at a Core haplotypes HLA-C2/C2 HLA-C1/C1 HLA-Bw4 high frequency and defines the second most common core haplotype at 17% in the Irish. Analysis of KIR2DL4 KIR3DL1*01502 ++ ÀÀ alleles showed a high frequency of functionally distinct KIR3DL1*00101 À + variants (9A, 10A-A and 10A-B alleles) present in the KIR3DS1 ÀÀ Irish. Similarly, KIR2DS4 alleles with different expression phenotypic characteristics are also present at high (b) frequencies. Thus, at least at the gene level, balancing Allelic components of Associations with HLA-C2/C2 KIR3DL1*01502 haplotype selection is keeping functionally distinct alleles of several polymorphic KIR loci at a high frequency in the Irish Total Male Female population. Evidence of natural selection is also observed at the KIR3DL1*01502 ++ NS +++ haplotype level. Analysis of LD and haplotype structure KIR2DL4*00102 ++ NS ++ KIR2DS4*00101 NS NS + in the Irish shows that polymorphisms at KIR2DL4, KIR2DL3*001 + NS NS KIR3DL1 and KIR2DS4 gene loci define haplotype cores, broadly similar to those defined in the Japanese popula- 18 Abbreviations: HLA, human leukocyte antigen; KIR, killer immuno- tion. This was despite major differences in the KIR globulin-like receptor; NS, not significant. genetic profiles of the two populations, which are Frequencies of KIR and HLA compound genotypes were tested for characterized by different allele spectra, for example, evidence of nonrandom association (see Table 2 for counts and neither KIR3DL1*004 nor KIR3DL1*002 alleles are pre- actual P-values) and a summary of the results are shown here. (a) sent in the Japanese,18 but are present at relatively high shows analysis for three core KIR haplotypes defined by 3DL1 frequencies in the Irish (see Figure 1). As individual polymorphism and (b) shows alleles present on the KIR haplotype component genes differ in terms of expression and defined by KIR3DL1*01502. ‘+’ indicates a positive association function, the KIR core haplotypes also differ in terms of (+Po0.05; ++Po0.02 and +++Po0.002) and ‘À‘ indicates a negative their functional capacity, for example, the Irish have one association (ÀPo0.05 and ÀÀPo0.01) between the genes. core (defined by KIR3DL1*004), which does not encode

Table 4 Gender differences in KIR:HLA compound genotype frequencies

Compound genotype Female freq. (n ¼ 115) Male freq. (n ¼ 122) P-value w2/FE

KIR2DL4 KIR2DL4*00102/3+/C2C2 0.080 (10) 0.070 (9) 0.7087 0.1396 KIR2DL4*00102/3+/C1C1 0.130 (15) 0.259 (29) 0.0338 4.5060 KIR2DL4*00102/3+/C1C2 0.165 (19) 0.197 (24) 0.5294 0.3956

Female freq. (n ¼ 115) Male freq. (n ¼ 121)

KIR3DL1 KIR3DL1*01502+/C2C2 0.061 (7) 0.025 (3) 0.2067 FE KIR3DL1*01502+/C1C1 0.061(7) 0.091 (11) 0.3848 0.7552 KIR3DL1*01502+/C1C2 0.061 (7) 0.041 (5) 0.4945 0.4668

Female freq. (n ¼ 114) Male freq. (n ¼ 119)

KIR2DS4 KIR2DS4*00101+/C2C2 0.061 (7) 0.059 (7) 0.9340 0.0068 KIR2DS4*00101+/C1C1 0.105 (12) 0.210 (25) 0.0286 4.7890 KIR2DS4*00101+/C1C2 0.140 (16) 0.160 (19) 0.9999 0.1701

Abbreviations: FE, Fisher’s exact test; freq., frequency; HLA, human leukocyte antigen; KIR, killer immunoglobulin-like receptor. The frequency at which HLA-C1 and HLA-C2 are found in combination with components of the KIR3DL1*01502 core haplotype were compared between males and females to assess the extent of gender-specific signatures of natural selection as shown in Table 2. Frequencies reflect combinations of KIR genotypes (that is, presence of a KIR) coupled with different HLA-C genotypes. The frequencies and counts (in brackets) are indicated. Differences between males and females were compared by w2 analysis or Fisher’s exact test where appropriate. Statistically significant P-values are indicated in bold (P-values o0.05). + indicates presence of a particular KIR allele; C1C1 or C2C2 indicate homozygosity for HLA-C1 or -C2, while C1C2 indicates heterozygosity for HLA-C1 and -C2.

Genes and Immunity KIR and HLA genes are coevolving KJ Guinan et al 475 any cell surface expression variants of the core genes. ‘activating’ core KIR haplotypes. Alternatively, our data This is likely to result in a phenotype that is very indicate that the core haplotype defined by different to other cores and that may have unique KIR3DL1*01502, or one of its component parts, is functional properties. Given the important role of KIR undergoing negative selection in the presence of the in NK cell-mediated innate immunity, it is likely that the weaker HLA-C1 ligand. The opposite HLA associations combined functional characteristics of polymorphisms are found for the strongly ‘inhibitory’ KIR3DL1*00101 within the core haplotype region have undergone natural core. We also found that the core haplotype defined by selection. We see evidence of this with balancing KIR3DS1 is negatively associated with its putative ligand selection maintaining diversity of KIR genes and HLA-Bw4, consistent with a number of studies that show haplotypes in the Irish, for example, the highest that KIR3DS1 is coevolving with its ligand.16,28 frequency haplotype was present at only 6.3% (compared The fact that the positive association of with 35.7% in the Japanese). Examination of the 25 most KIR3DL1*01502:HLA-C2 is female specific is strongly common haplotypes in the Irish shows only four in suggestive of a selective mechanism involving NK cell common with the Japanese. The most common core function at the level of human reproduction. For haplotype in the Japanese (defined by KIR3DL1*01502 combinations of polymorphisms such as found in KIR and present at a frequency of 43.1%) is also found at a and HLA genes to be over- or underrepresented in a relatively high frequency in the Irish (10.6%, see particular gender suggests a selection mechanism that is Figure 3a). The high frequency and high homozygosity most likely to promote survival either during pregnancy extending from both ends of this core haplotype in the (sex-specific prenatal selection) or after pregnancy (for Irish combine to give a signature of positive selection, example, gender-specific diseases).18,30 During fetal similar to that observed in the Japanese.18 Thus, at least development, male embryos exhibit greater uterine for KIR3DL1*01502, there seems to be evolutionary- vulnerability than females giving rise to a higher male selective forces operating at the haplotype level in to female ratio at fertilization (X1.20:1) than at birth different human populations. (1.06:1).31,32 There is also evidence of specific develop- Analysis of LD and haplotype structure can also be mental disadvantages, which lead to higher rates of informative in the design of disease association studies. spontaneous abortions in female embryos during early Although highly polymorphic, the allele present at the gestation.33 Any fetal characteristic that could circum- KIR3DL1 locus in the Irish population can be used for vent gender differences in survival is likely to be haplotype tagging as it can uniquely define the alleles overrepresented in future populations of males or present at both KIR2DL4 and KIR2DS4 loci. Indeed, the females.18 Genetic factors may also be selected for during seven most common cores that account for almost 95% of pregnancy as they may contribute to altered uterine NK the core haplotypes in the Irish population can be cell function and reproductive success.30 However, identified exclusively by the allele at KIR3DL1. Haplo- selective pressures acting on KIR and HLA genes may type tagging in the Irish population can also be extended also occur after birth.18 It is becoming apparent that most outside the core haplotype for particular combinations human diseases exhibit differences between the gen- of alleles with the allele present at KIR3DL1 (for ders34 and this can potentially affect the relative KIR3DL1*002, *01502 and KIR3DS1) indicating the reproductive potential of males and females, for exam- allele present at KIR3DL2. This is hugely informative ple, morbidity associated with a particular female- and can be used as a basis for disease association studies associated disease may preclude pregnancy. Thus, and clinical studies such as defining the role of KIR genes although evidence is growing to suggest that gender- in transplantation in this well-characterized Irish specific immune-related selective pressures occur, their population. mechanisms of action remain unknown. Similar to KIR genes, HLA class I genes are highly It is striking that some of the observed KIR:HLA polymorphic and rapidly evolving. It is thought that associations in the Irish population directly contrast diverse HLA class I ligands may influence evolution of with those found in the Japanese, for example, a NK cell receptors as their interactions are functional.18 positive nonrandom association was found between Although KIR and HLA class I genes segregate indepen- KIR3DL1*01502 and HLA-C1 in Japanese females.18 dently on different chromosomes, analysis of KIR:HLA Although there is the possibility for type 1 statistical receptor ligand combinations in the Irish population has errors that only analysis of larger cohorts can resolve, revealed several nonrandom associations that indicates differences in particular KIR:HLA associations, including that they are coevolving. Evidence for natural selection those that are gender specific, might imply that selective should be considered in context of the functional pressures acting on KIR genes may be different in both characteristics conferred by components of KIR haplo- populations. This may not be surprising considering that types,18 and the importance of maintaining a balance both are genetically homogeneous, relatively isolated between inhibitory and activating signals in controlling island populations, they belong to different ethnic NK cell function. Unlike others, the core haplotype groups and have contrasting pathogenic histories that defined by KIR3DL1*01502 (moderate inhibitory recep- are likely to have shaped their HLA and KIR repertoires. tor) contains fully expressed and functional KIR2DL4 Unlike the Japanese population, B-haplotype KIR and KIR2DS4 alleles, which renders this a haplotype that frequencies are much higher in the Irish and it is is more ‘activating’.18 This haplotype is found to very common for individuals to inherit B-haplotypes positively associate with HLA-C2 and/or to negatively in a heterozygous state along with A haplotypes, such associate with HLA-C1. In functional terms, HLA-C2 is as those defined by KIR3DL1*01502. Consequently, in considered a strong inhibitory ligand and the association the context of additional B-haplotype activating KIR between the two is likely to reflect a selective pressure being present, the functional impact of carrying the acting to maximize inhibition in individuals with KIR3DL1*01502 haplotype in the Irish population is

Genes and Immunity KIR and HLA genes are coevolving KJ Guinan et al 476 likely to be considerably different than in the Japanese. approach based on the method of van Oosterhout Indeed, a recent finding shows that NK cells from et al.51 (and C van Oosterhout,qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi personal communication); individuals heterozygous for the KIR B-haplotype are solve for allele x: Fx ¼ ðF ÞÀð2FxFnullÞ. more responsive to pathogen-associated signals than are observed homozygotes homozygotes.35 Thus, contrasting signatures of KIR and HLA coevolution in the Irish and Japanese populations Natural selection may reflect, at least in part, the functional impact of Departure from neutrality of KIR allele frequencies having contrasting B-haplotype KIR repertoires in these was examined using the Ewens Watterson statistic, populations. Our data also suggest that it is unlikely that implemented by the BOTTLENECK program.19 Using any particular KIR and HLA combination reflects a KIR allele frequency data, the program calculates and fundamental trait to benefit human reproduction in compares observed heterozygosity with levels expected general, but the finding that the core haplotype defined under neutrality according to the infinite allele model. by KIR3DL1*01502 has a strong association with HLA Higher than expected diversity is consistent with a only in females in two diverse populations suggests that history of diversifying/balancing selection acting at a it may indeed have some role in human reproduction, locus, which gives rise to several variants occurring at the nature of which remains to be elucidated. high frequency in a population. Lower than expected diversity is indicative of purifying/directional selection acting to elevate a single allele to high frequency in a Materials and methods population. Analysis of LD between alleles of several polymorphic Subjects and DNA isolation KIR genes, all of which were in Hardy–Weinberg Blood donors (n ¼ 240) were recruited from the Blood equilibrium (KIR2DL3, KIR3DL1, KIR2DL4, KIR2DS4 Transfusion Services of Belfast City Hospital, Belfast, and KIR3DL2), was carried out using the MIDAS pro- Northern Ireland, UK, and St James’ Hospital Dublin, gram.20 From individual genotype data, D0 values were Republic of Ireland. Ethics approval was received from calculated through the pairwise comparison of each allele the appropriate committees in both centers and written combination defining a two-locus haplotype. Significance consent was obtained from all donors. Mononuclear cells of D0 values was determined by w2 analysis using a Yates were isolated from buffy coats using standard Ficoll correction. The graphical output was annotated and gradient centrifugation. Genomic DNA (gDNA) was modified for presentation using Graphic converter extracted from cells using the salting-out method.36 (Version 5.2, Lemke Software, Peine, Germany; www. lemkesoft.com). KIR gene and allele typing In the absence of segregation analysis, algorithms can PCR-SSOP for KIR genes was performed as previously 37,38 be used to reconstruct haplotypes in a population from described. For KIR allele subtyping, PCR amplifica- individual genotyping information. In this study, haplo- tions and SSOP hybridizations were performed as types were reconstructed at the allele level using a previously described: KIR2DL3,39 KIR3DL2,40 KIR2DS4,41 42 43 44 Bayesian statistical method implemented by the PHASE KIR2DL4, KIR3DL1/S1, KIR2DS1, KIR2DL5 and 21,22 45 program (version 2.1, ). On the basis of LD and known KIR2DS5. Upgraded allele typing systems were devised haplotype structure within the KIR gene complex, a for KIR2DS4 (to distinguish *003 from *007), KIR2DL4 (to number of constraints were used in haplotype estima- discriminate between 9A and 10A alleles) and to resolve tion: (1) KIR2DL2 is an allele of KIR2DL3; (2) KIR3DS1 is KIR3DL2 subtyping of genotypes containing ambiguous 46 an allele of KIR3DL1; (3) KIR3DL1, KIR3DL2, KIR3DL3, combinations of certain alleles. Only one allele of KIR3DP1, KIR2DL3 and KIR2DL4 are always present; (4) KIR2DS1 and KIR2DS5 was present in the Irish. in contrast, positive PCR for KIR2DL1, KIR2DP1, KIR2DL5A or KIR2DL5B genes were distinguished by KIR2DS1, KIR2DS2, KIR2DS3, KIR2DS4, KIR2DS5, KIR2- the typing. KIR genes and frequencies were almost DL5A and KIR2DL5B does not confirm homozygosity, identical between the groups from Northern Ireland and but the presence of at least one copy of the gene/allele. the Republic of Ireland (data not shown) and the data Full-length haplotypes containing 16 KIR genes were were combined. The HLA class I types of the donors reconstructed. were defined using a previously described SSOP 47 Evidence of natural selection can also be detected by method. The following ligands were distinguished: investigating haplotype homozygosity.24 In this type of HLA-C1, HLA-C2 and HLA-B Bw4. Sex determination 48 analysis, core haplotype regions (in our data set, was performed by PCR of a zinc-finger protein gene. consisting of three KIR genes) are defined by unique combination of alleles that they have at these loci. Statistical analysis Examination of diversity at loci outside the core region KIR allele frequencies (F) were determined using the can provide information on whether components of an formula F ¼ Fhomozygotes þ (Fheterozygates/2) for high-fre- extended haplotype have been selected for. A haplotype quency KIR genes, KIR2DL3, KIR2DL4, KIR3DL1 and recently selected for and driven to high frequency in a KIR3DL2.49 In case of KIR2DS4, the frequency of the population is likely to display high homozygosity at null allele type (that is, absence of the gene) was genes on both sides of the core region. Haplotype estimated using the frequency of null homozygotes homozygosity was investigated according to the method accordingpffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi to the equation of Bernstein et al.:50 of Sabeti et al.,24 using a web-based ‘Extended Haplotype 52 Fnull ¼ Observed null homozygote frequency. Frequen- Homozygosity’ tool. Core regions were defined as areas cies of individual KIR2DS4 alleles were then calculated of high LD encompassing KIR2DL4, KIR2DS4 and based on their observed homozygosity frequencies and KIR3DL1, on the basis of the statistical output derived the frequency estimate of the null allele using an from LD analysis and with reference to the findings of

Genes and Immunity KIR and HLA genes are coevolving KJ Guinan et al 477 Yawata et al.18 Haplotypes that were identified in one biased toward recognition of HLA-C and alters with gesta- or more donors were included in further analysis. tional age. J Immunol 2008; 181: 39–46. These haplotypes occurred at a combined frequency of 9 Trowsdale J, Moffett A. NK receptor interactions with MHC 85.4% in our population and included rare singletons at a class I molecules in pregnancy. Semin Immunol 2008; 20: 317–320. combined frequency of 3.5%. Haplotype homozygosity 10 Martin MP, Single RM, Wilson MJ, Trowsdale J, Carrington M. was measured at centromeric and telomeric ends of KIR haplotypes defined by segregation analysis in 59 Centre the core, on the basis of polymorphism at KIR2DL3 d’Etude Polymorphisme Humain (CEPH) families. Immuno- and KIR3DL2, respectively. Haplotype homozygosity genetics 2008; 60: 767–774. was represented graphically using GraphPad Prism 11 Hsu KC, Chida S, Geraghty DE, Dupont B. The killer cell immunoglobulin-like receptor (KIR) genomic region: gene- (Version 4.0.a for MAC, GraphPAd Prism Software, order, haplotypes and allelic polymorphism. Immunol Rev San Diego, CA, USA; www.graphpad.com). Values were 2002; 190: 40–52. also plotted against core haplotype frequency to view the 12 Trowsdale J, Barten R, Haude A, Stewart CA, Beck S, Wilson relationship between haplotype diversity and frequency. MJ. The genomic context of natural killer receptor extended Haplotype homozygosity values and frequencies of gene families. Immunol Rev 2001; 181: 20–38. a nonevolving, neutral haplotype data set were derived 13 Ahlenstiel G, Martin MP, Gao X, Carrington M, Rehermann B. from Mueller and Andreoli52 and used for general Distinct KIR/HLA compound genotypes affect the kinetics of comparison. human antiviral natural killer cell responses. J Clin Invest 2008; Analysis of KIR and HLA associations in the popula- 118: 1017–1026. tion involved a systematic comparison of the presence 14 Alter G, Martin MP, Teigen N, Carr WH, Suscovich TJ, of KIR genes, genotypes and alleles with KIR ligands Schneidewind A et al. Differential natural killer cell-mediated encoded by HLA class I, specifically HLA-C1, HLA-C2 inhibition of HIV-1 replication based on distinct KIR/HLA 204 and HLA-Bw4. The potential for nonrandom KIR–HLA subtypes. J Exp Med 2007; : 3027–3036. 15 Kim S, Sunwoo JB, Yang L, Choi T, Song YJ, French AR et al. w2 associations was assessed by and Fisher’s exact tests. HLA alleles determine differences in human natural killer cell responsiveness and potency. Proc Natl Acad Sci USA 2008; 105: 3053–3058. 16 Single RM, Martin MP, Gao X, Meyer D, Yeager M, Kidd JR Conflict of interest et al. Global diversity and evidence for coevolution of KIR and The authors declare no conflict of interest. HLA. Nat Genet 2007; 39: 1114–1119. 17 Guinan K, Cunningham RT, Meenagh A, Dring MM, Middleton D, Gardiner CM. Receptor systems controlling natural killer cell function are genetically stratified in Europe. Acknowledgements Genes Immun 2010; 11: 67–78. 18 Yawata M, Yawata N, Draghi M, Little AM, Partheniou F, This work was supported by an Ireland–Northern Parham P. Roles for HLA and KIR polymorphisms in natural Ireland Co-operation grant funded jointly by the Irish killer cell repertoire selection and modulation of effector Health Research Board and the Northern Ireland R&D function. J Exp Med 2006; 203: 633–645. Office. CMG is also supported by the Science Foundation 19 Cornuet JM, Luikart G. Description and power analysis of two Ireland. We also acknowledge Dr Anthony W Ryan, Dr tests for detecting recent population bottlenecks from allele 144 Pardis D Sabeti, Dr Aisling M O’Halloran, Dr Cock van frequency data. Genetics 1996; : 2001–2014. 20 Gaunt TR, Rodriguez S, Zapata C, Day IN. MIDAS: software Oosterhout, Maria Morrison and Dr Brian McSharry for for analysis and visualisation of interallelic disequilibrium their helpful discussions about the study. between multiallelic markers. BMC Bioinformatics 2006; 7: 227. 21 Stephens M, Donnelly P. A comparison of bayesian methods for haplotype reconstruction from population genotype data. Am J Hum Genet 2003; 73: 1162–1169. References 22 Stephens M, Smith NJ, Donnelly P. A new statistical method for haplotype reconstruction from population data. Am J Hum 1 Parham P. The genetic and evolutionary balances in human Genet 2001; 68: 978–989. NK cell receptor diversity. Semin Immunol 2008; 20: 311–316. 23 Middleton D, Meenagh A, Gourraud PA. KIR haplotype 2 Hanna J, Goldman-Wohl D, Hamani Y, Avraham I, Greenfield C, Natanson-Yaron S et al. Decidual NK cells regulate key content at the allele level in 77 Northern Irish families. developmental processes at the human fetal-maternal inter- Immunogenetics 2007; 59: 145–158. face. Nat Med 2006; 12: 1065–1074. 24 Sabeti PC, Reich DE, Higgins JM, Levine HZ, Richter DJ, 3 O’Connor GM, Hart OM, Gardiner CM. Putting the natural Schaffner SF et al. Detecting recent positive selection in the killer cell in its place. Immunology 2006; 117: 1–10. human genome from haplotype structure. Nature 2002; 419: 4 Gardiner CM. Killer cell immunoglobulin-like receptors on NK 832–837. cells: the how, where and why. Int J Immunogenet 2008; 35: 1–8. 25 Martin AM, Kulski JK, Gaudieri S, Witt CS, Freitas EM, 5 Lanier LL. Activating and inhibitory NK cell receptors. Adv Trowsdale J et al. Comparative genomic analysis, diversity Exp Med Biol 1998; 452: 13–18. and evolution of two KIR haplotypes A and B. Gene 2004; 335: 6 Hiby SE, Regan L, Lo W, Farrell L, Carrington M, Moffett A. 121–131. Association of maternal killer-cell immunoglobulin-like re- 26 Yawata M, Yawata N, Abi-Rached L, Parham P. Variation ceptors and parental HLA-C genotypes with recurrent within the human killer cell immunoglobulin-like receptor miscarriage. Hum Reprod 2008; 23: 972–976. (KIR) gene family. Crit Rev Immunol 2002; 22: 463–482. 7 Hiby SE, Walker JJ, O’Shaughnessy KM, Redman CW, 27 Gutierrez-Rodriguez ME, Sandoval-Ramirez L, Diaz-Flores M, Carrington M, Trowsdale J et al. Combinations of maternal Marsh SG, Valladares-Salgado A, Madrigal JA et al. KIR gene KIR and fetal HLA-C genes influence the risk of preeclampsia in ethnic and Mestizo populations from Mexico. Hum Immunol and reproductive success. J Exp Med 2004; 200: 957–965. 2006; 67: 85–93. 8 Sharkey AM, Gardner L, Hiby S, Farrell L, Apps R, Masters L 28 Norman PJ, Abi-Rached L, Gendzekhadze K, Korbel D, et al. Killer Ig-like receptor expression in uterine NK cells is Gleimer M, Rowley D et al. Unusual selection on the

Genes and Immunity KIR and HLA genes are coevolving KJ Guinan et al 478 KIR3DL1/S1 natural killer cell receptor in Africans. Nat Genet 41 Maxwell LD, Williams F, Gilmore P, Meenagh A, Middleton D. 2007; 39: 1092–1099. Investigation of killer cell immunoglobulin-like receptor gene 29 Ferrer-Admetlla A, Bosch E, Sikora M, Marques-Bonet T, diversity: II. KIR2DS4. Hum Immunol 2004; 65: 613–621. Ramirez-Soriano A, Muntasell A et al. Balancing selection is 42 Williams F, Meenagh A, Sleator C, Middleton D. Investigation the main force shaping the evolution of innate immunity of killer cell immunoglobulin-like receptor gene diversity: I. genes. J Immunol 2008; 181: 1315–1322. KIR2DL4. Hum Immunol 2004; 65: 31–38. 30 Ucisik-Akkaya E, Dorak MT. A study of natural killer cell 43 Halfpenny IA, Middleton D, Barnett YA, Williams F. Investi- lectin-like receptor K1 gene (KLRK1/NKG2D) region poly- gation of killer cell immunoglobulin-like receptor gene morphisms in a European population sample. Tissue Antigens diversity: IV. KIR3DL1/S1. Hum Immunol 2004; 65: 602–612. 2009; 73: 177–183. 44 Meenagh A, Gonzalez A, Sleator C, McQuaid S, Middleton D. 31 McMillen MM. Differential mortality by sex in fetal and Investigation of killer cell immunoglobulin-like receptor gene neonatal deaths. Science 1979; 204: 89–91. 32 Vatten LJ, Skjaerven R. Offspring sex and pregnancy outcome diversity, KIR2DL1 and KIR2DS1. Tissue Antigens 2008; 72: by length of gestation. Early Hum Dev 2004; 76: 47–54. 383–391. 33 Eiben B, Bartels I, Bahr-Porsch S, Borgmann S, Gatz G, Gellert 45 Gonzalez A, Meenagh A, Sleator C, Middleton D. Investiga- G et al. Cytogenetic analysis of 750 spontaneous abortions tion of killer cell immunoglobulin-like receptor (KIR) gene with the direct-preparation method of chorionic villi and its diversity: KIR2DL2, KIR2DL5 and KIR2DS5. Tissue Antigens implications for studying genetic causes of pregnancy wa- 2008; 72: 11–20. stage. Am J Hum Genet 1990; 47: 656–663. 46 Gonzalez A, McErlean C, Meenagh A, Shovlin T, Middleton D. 34 Ober C, Loisel DA, Gilad Y. Sex-specific genetic architecture of Killer cell immunoglobulin-like receptor allele discrimination human disease. Nat Rev Genet 2008; 9: 911–922. by high-resolution melting. Hum Immunol 2009; 70: 858–863. 35 Korbel DS, Norman PJ, Newman KC, Horowitz A, Gendze- 47 Middleton D, Halfpenny I, Meenagh A, Williams F, Sivula J, khadze K, Parham P et al. Killer Ig-like receptor (KIR) Tuomilehto-Wolf E. Investigation of KIR gene frequencies in genotype predicts the capacity of human KIR-positive type 1 diabetes mellitus. Hum Immunol 2006; 67: 986–990. CD56dim NK cells to respond to pathogen-associated signals. 48 Fredsted T, Villessen P. Fast and reliable sexing of prosimian J Immunol 2009; 182: 6426–6434. and human DNA. Am J Primatol 2004; 64: 345–350. 36 Miller SA, Dykes DD, Polesky HF. A simple salting out 49 Halliburton R. Quantifying genetic variation. In: T RC (ed). procedure for extracting DNA from human nucleated cells. Introduction to Population Genetics, vol. International edn. Nucleic Acids Res 1988; 16: 1215. Pearson Education, Inc.: Upper Saddle River, 2004. 37 Middleton D, Curran M, Maxwell L. Natural killer cells and 50 Bernstein F. Uber die Erblichkeit der Blutgruppen. Zeitschrift their receptors. Transpl Immunol 2002; 10: 147–164. 38 Middleton D, Williams F, Halfpenny IA. KIR genes. Transpl fu¨r induktive Abstammungs und Vererbungslehre. 1930; 54: Immunol 2005; 14: 135–142. 400–426. 39 Keaney L, Williams F, Meenagh A, Sleator C, Middleton D. 51 Van Oosterhout C, Weetman D, Hutchinson WF. Estimation Investigation of killer cell immunoglobulin-like receptor gene and adjustment of microsatellite null alleles in nonequilibrium diversity III. KIR2DL3. Tissue Antigens 2004; 64: 188–194. populations. Molecular Ecology Notes 2006; 6: 255–256. 40 Meenagh A, Williams F, Sleator C, Halfpenny IA, Middleton 52 Mueller JC, Andreoli C. Plotting haplotype-specific linkage D. Investigation of killer cell immunoglobulin-like receptor disequilibrium patterns by extended haplotype homozygosity. gene diversity V. KIR3DL2. Tissue Antigens 2004; 64: 226–234. Bioinformatics 2004; 20: 786–787.

Genes and Immunity