Complex interactions: The immunogenetics of human leukocyte antigen and killer cell immunoglobulin-like receptors

Complex Interactions: The Immunogenetics of Human Leukocyte Antigen and Killer Cell Immunoglobulin-like Receptors Paul J. Norman and Peter Parham The killer cell immunoglobulin-like receptors (KIR) for human leukocyte antigen (HLA) modulate innate and adaptive immunity by controlling effector cells. HLA and KIR are encoded in genomic regions that have complex organization and exhibit exceptional diversity within and among human population groups. This diversity is likely to have arisen to combat a constantly evolving pathogen challenge. Numerous variations influence the expression level or function of KIR molecules and can affect their interaction with HLA, with important implications for the immune response. The functional variety of natural immune responses that are controlled by HLA and KIR interactions is genetically determined and maintained by natural selection. Semin Hematol 42:65-75 © 2005 Elsevier Inc. All rights reserved.

N

atural killer (NK) cells migrate from the blood in response to danger signals, where their interaction with tissue cells via histocompatibility antigens helps control the development of immune reactions. Among the impressive arsenal of receptor/ligand interactions that regulate this innate immune response,1,2 the relationship between human leukocyte antigen (HLA) class I and the killer cell immunoglobulin-like receptors (KIR) molecules is one of the most important.3,4 HLA and KIR gene families have complex genetic variation, organization, and expression patterns, which contribute towards individual variation in immunity.5 These genetically prescribed differences affect the function of either molecule independently, or their interaction.6,7 As HLA and KIR are encoded by separate chromosomes, but neither is subject to somatic development; randomly inherited combinations of receptor and ligand can determine capability to combat disease or to tolerate transplantation. By their very definition, NK cells were recognized as potential therapeutic agents able to kill infected or transformed cells without prior activation, while inducing and perpetuating the inflammatory response with cytokines.8 Through a

Departments of Structural Biology and Microbiology and Immunology, Stanford University School of Medicine, Stanford, CA. Supported in part by a postdoctoral fellowship from the Lymphoma Research Foundation (P.J.N.) and a grant awarded by the Leukaemia and Lymphoma Society of the USA. Address correspondence to Paul J. Norman, PhD, Sherman Fairchild Building D159, 299 Campus Dr W, Stanford University School of Medicine, Stanford, CA 94305-5126. E-mail: [email protected]

0037-1963/05/$-see front matter © 2005 Elsevier Inc. All rights reserved. doi:10.1053/j.seminhematol.2005.01.007

balance of inhibitory and activating signals, NK cells respond to changes in HLA expression, which may be naturally provoked by some micro-organisms or during tumorigenesis.9,10 As class I molecules interact with the T-cell receptor (TCR) of cytotoxic T cells in addition to KIR, both the quantity and quality of HLA molecules can be monitored by the cells of adaptive and innate immunity. This link between the two broad arms of the immune response is reinforced when NK cells interact with adaptive immune mediators that may also be present at sites of infection.11,12 To counter the large and changing array of pathogens requires variation in immune response molecules.4 For effector molecules of adaptive immunity, the required variety is achieved by somatic rearrangement of genomic DNA and selection for appropriately acting cells. For HLA and KIR molecules, functional variation is determined by their substantial genetic polymorphism.5,13 Here we describe the specific immunologic implications of genetic variation for HLA and particularly KIR. Directly or not, genetic variation begets variation in quantity, distribution, or function of protein expression (or phenotype). Such variations are crucial for the interaction of HLA and KIR, the immune response, and ultimately for tissue transplantation.

Polymorphism and Genetic Organization Polymorphic HLA molecules are encoded in the major histocompatibility complex (MHC) of chromosome 6p21.3 and 65

P.J. Norman and P. Parham

66

Figure 1 The KIR gene cluster. Each KIR has two or three immunoglobulin domains and a long (L) or short (S) cytoplasmic tail that signals inhibition or activation, respectively.10 P ⫽ pseudogene. KIR2DL1-3 and 2DS1-5 bind HLA-Cw, 3DL1 binds HLA-B allotypes, and 3DL2 binds HLA-A allotypes. Framework loci that are found on most haplotypes are shown as larger boxes. 3DL1 and 3DS1 are alleles. 3DP1-2DL4-3DL/S1 constitute central hub with component haplotypes either side; width of line corresponds to frequency in Caucasians.38 Not all possible haplotypes are depicted.21 The most frequent haplotype (A; follow the thickest line from 3DL3 to 3DL2) encodes only 2DS4 as a potential activating receptor. Most other genotypes are termed “B”. The somatic ratio of inhibitory to activating receptors is determined by haplotype.

KIR in the leukocyte receptor complex (LRC) of chromosome 19q13.4. For the purposes of this review, we define genetic polymorphism as any difference in DNA sequence that can be inherited, and an allele as the entire genomic sequence (locus) that includes all of the protein-coding information and surrounding genetic material. We use the immunogeneticist’s interpretation of the term “haplotype,” which refers to coinheritance of complete alleles on a chromosome (otherwise an allele or a haplotype may refer to any unit of DNA from 1 base pair upwards); it is not usually established whether common haplotypes share identity at intermediate loci.

HLA Subtypes of HLA class I-A and -B and probably all -Cw molecules interact with KIR. Approximately 250 alleles of HLA-A, 500 of -B, and 120 of -C are known (for regular updates, see www.ebi.ac.uk/imgt/hla). Their allelic diversity is formed by point mutations that have been shuffled by many and varied DNA recombination events, including gene conversion.13,14 The MHC region (sometimes called the HLA region) encodes more than 200 polypeptides, up to 40% of which are involved in the immune response, and spans approximately 4 Mb (mega base pairs), with the class I region accounting for about 2 Mb.4,13 HLA class I molecules expose peptide fragments obtained from intracellularly generated molecules to the TCR, enabling appropriately adapted and primed cytotoxic T cells to identify and eliminate cells expressing foreign proteins. The polymorphic residues that define HLA variation cluster around regions influencing peptide contact, so that the various alleles are able to bind different arrays of peptides and thus combat a wide range of pathogens. Polymorphic residues also contribute to the TCR interface, which can result in a more varied and efficient TCR repertoire in heterozygous individuals.15 Thus, class I polymorphism helps insure our mutual immunity against a range of pathogens by defining variation within and between individuals.13,16

LRC and KIR Functional KIR diversity is determined by allelic polymorphism that is superimposed on variation in gene number and content, and by variegated KIR expression within the individual.4,5 The seven to 15 loci are clustered within 65 to 200 kb, where each KIR molecule is encoded by a separate KIR gene (Fig 1). KIR isotypes are defined by number of immunoglobulin subunits and their potential to inhibit or stimulate cellular activity. NK cells are regulated by a balance of these signals, most importantly the inhibitory signals, which dominate over-activation by KIR and other receptors.9,10 The LRC also encodes leukocyte immunoglobulin-like receptors (LILR; the receptors previously known as ILT, LIR, or MIR), some of which may be important during the KIR/HLA interaction.17–19 New KIR genes have been generated by meiotic recombination that has shuffled the exons\domains of existing KIR as entire units.20 Likewise, new KIR haplotypes have been formed by duplication, deletion, and reassortment around a framework hub containing the 3DP1, 2DL4, and 3DL1 loci.21 The result has been an accumulation of KIR haplotypes distinguished by gene content that vary also in their dispersal among contemporary human population groups.22,23 The simplest known haplotype is also the most common (haplotye A; Fig 1) but is not well represented in some ethnic groups (Fig 2), while the majority of gene-content diversity in all populations arises due to the many uncommon haplotypes. “New” HLA alleles are continually being encountered, but their rate of discovery has probably peaked. However, the extent of allelic diversity is not known for any of the KIR, although those KIR that interact with the two most polymorphic HLA class I molecules have the largest number of characterized alleles.23 Variation throughout the KIR molecule suggests that factors other than direct HLA interaction have been involved in their evolution.20 Activating KIR can bind HLA but may also recognize other molecules such as viral proteins. Mice have a family of receptors with similar genomic characteristics and function to KIR, called Ly49.2 Murine cytomegalovirus can cause infected cells to display

Immunogenetics of HLA and KIR

67

Figure 2 Prevalence of “A” haplotype and activating KIR varies between populations. Results are reviewed elsewhere.22,23 All individuals were genotyped for presence/absence variation of each KIR. Populations ordered by A haplotype frequency (shown as %); the B haplotypes are more diverse (Fig 1) so that no single haplotype forms the reciprocal proportion. There have been more than 20 B haplotypes described in Caucasians alone, but the diversity of genotypes in other populations indicates many more.31

molecular mimics of murine class I molecules, thus protecting them from NK cells that display inhibitory Ly49. Activating Ly49, which are not found in all mouse strains, have probably evolved to counteract this attempted evasion.2

Genetic Polymorphism Affects Protein Expression NK cell activity can be influenced by the cell surface density of either HLA or KIR6; genetic variation affects the expression level of either molecule. A nonsynonymous DNA variation changes the resulting amino acid; synonymous variants incur no direct coding change but may influence mRNA splicing,24 expression quantity,5 and tissue distribution.25

Isotypic Variation HLA-A, -B, and -Cw molecules are present on the surface of most nucleated cells. HLA-Cw is normally at a lower level than -A or -B, which may enhance the sensitivity for some NK cells to detect changes in its expression. Aside from modest influence by HLA,5 no known polymorphic factors control KIR expression, so the functional KIR repertoire is directly dependent on KIR genes, the first tier of variation being presence or absence of the gene. Assuming that the locus is present, expression of most KIR molecules is stochastically acquired and results in a highly heterogeneous NK cell population.5 This variegated repertoire allows the NK cell population as a whole to respond to changes in the cell surface

density of any class I molecule.6,10 Once KIR expression is established it remains stable,5 which may equip NK cells with a reference that enables sensitivity to changes in HLA levels. KIR are also expressed by T cells that display memory phenotype, suggesting a more regulated mode of acquisition in this case.26 NK cells require at least one inhibitory molecule to develop, and when their randomly acquired expression results in cells without KIR, other inhibitory receptors can compensate.1,10 The original observation that KIR are expressed (or not), regardless of their coexisting HLA ligand, was confirmed recently when a transgenic mouse was generated using a complete KIR cluster: the mouse expressed KIR despite complete lack of natural ligands.4 The cytosine residues from CG dinucleotides (CpG islands) that are located in the DNA sequence upstream from some mammalian genes can become methylated and limit gene expression. Individual promoter regions are methylated in silent KIR genes and unmethylated in those expressed, deand re-methylation acting as a switch in vitro.4 During development, KIR genes may be made available for transcription by sequentially unraveling their DNA tertiary structure and de-methylation of developmentally dormant genes.26 For example, mRNA from 3DL3 is not detectable in peripheral blood NK cells and its promoter is methylated, implying that the gene is switched off at a certain stage of development or inducible in some cell populations25; indeed, once “switched on,” this KIR has a very high expression level.25 The mechanisms that drive this KIR expression may be conserved in

68 evolution, as the distribution of expression in KIR-transgenic mice is remarkably similar to that of humans.4 All KIR promoters contain a potential binding site for acute myeloid leukemia factor (AML or CBF; used here to distinguish it from the disease). There are three types of promoter; those for 2DL4 and 3DL3 are distinct from each other and the remaining KIR.4 This difference corresponds to their tissue distribution pattern: 2DL4 is expressed by particular subsets of peripheral NK cells,9,27 3DL3 by uterine NK cells,28 and the remaining KIR randomly assorted among peripheral NK cells. CBF-A1 is active in all three types as a general repressor of DNA transcription in NK cells, but subtle differences in nontranslated DNA sequence are required to modulate KIRspecific expression.25 For example, 2DL4 is regulated by the basic CBF promoter and a combination of sequences unique to this locus, including nucleotide motifs both upstream and downstream of the translation-initiation codon that enable inhibition or enhancement of transcription, respectively.25 Detecting transcription-suppression by CBF-A1 contrasts with observations that natural mutations of the CBF binding site correlate with lack of KIR expression by peripheral blood NK cells.5 This inconsistency could have arisen because the isoforms of CBF (A1-3) have different effects and may be differentially expressed during NK cell development.25 Moreover, the potential for control of KIR expression by this DNA binding protein—mutated by chromosome 21q22 translocation in up to 20% AML cases–may implicate KIR in pathogenesis or as a phenotypic marker for disease.

Allotypic Variation Nucleotide substitution (or single-nucleotide polymorphism [SNP]) is the most basic unit of genetic variation. Nonsynonymous SNP or other mutations may introduce a configuration change or premature stop codon and prevent normal expression. There are several examples of these naturally occurring “knockout” and “knockdown” alleles of HLA. There are approximately 20 null alleles each of HLA-A and –B13 (how many HLA-Cw alleles are null is unknown because these variants have been mostly described by DNA analysis). Genetic mutations may cause a molecule to be secreted rather than expressed on the cell surface, as observed for a variant of the B44 allele.13 Synonymous polymorphism can influence mRNA stability, as observed during upregulation of the MHC-encoded corneodesmosin.24 Intron mutations may also occur near sites that control mRNA splicing and thus enhance or impair expression; for example, there are two such variants of A24 with low expression.13 Mutations in other MHC genes, such as TAP, can seriously hinder class I expression (TAP is the transporter associated with antigen processing; the protein product helps provide class I with peptide fragments). Such genetically determined restraint may have a more substantial influence on KIR than on HLA. A KIR gene may not be expressed at the cell surface, as nonexpressed variants have been found for many KIR, including 2DS4, 2DL4, 2DL5, 3DL1, and 3DP1. There is a common allele of 2DS4 characterized by a 22-bp deletion21 that introduces a premature stop codon and

P.J. Norman and P. Parham renders its mRNA transcript unlikely to be translated into a functional protein.29 Two common alleles of 2DL4 in Caucasians are distinguished by a 9 or 10A segment of coding sequence in the transmembrane domain-coding exon. The 9A allele has a premature stop codon and does not produce a functional protein.27 2DL5A and 2DL5B are distinct loci (Fig 1) that potentially encode very similar molecules. An asymmetric recombination has placed 2DL5B under control of a nonfunctional promoter from pseudogene 3DP1 and so 2DL5B is not expressed.5 KIR3DL1 allotypes may be distinguished according to their expression levels, and the residues with the most profound effect have been identified by reciprocating the substitutions that define the two extremes of this phenotype.30 3DL1*002 is observed at a relatively high cell surface level, while 3DL1*004 is not detectable. Nonsynonymous substitutions additively affected 3DL1 expression and one (SNP c319t; amino acid S086L) resulted in cellular retention of the molecule. Lastly, 3DP1 forms part of the central framework hub and is considered a pseudogene, due to lack of expression in peripheral blood. KIR3DP1 is not expressed because it contains a mutation in the CBF binding site,5 but in some individuals (about 4% of Caucasians) this KIR is expressed as a potentially secreted molecule.29 In what appears to be a role reversal of the situation observed for 2DL5B, the functioning promoter from 2DL5A has apparently recombined with 3DP1 to enable 3DP1 expression.5 HLA class I null homozygous individuals are rarely observed, but only two KIR have been detected in every individual that has been genotyped (3DL2 and 3DL3).22,31 Some KIR may be particularly prone to loss or changes incurred by meiotic recombination events because they have similar promoter sequences,4 as seen for 2DL5 and 3DP1. Null or lowexpression alleles of HLA and KIR may have prevailed as countermeasures to immune evasion by particular pathogens. A KIR-null allele will have the same impact as absence of the gene and may indicate that it is a disadvantage to possess some KIR when they are not currently required (or able) to help control infection. In contrast, low-expression KIR alleles may make the NK cell more sensitive to physiologic changes in HLA expression levels.6 HLA-null alleles have implications for tissue transplantation, such as possible NK-mediated graft-versus-host disease in HLA-null recipients of bone marrow, as well as potential benefits when allocating solid organs. Their discovery is partially responsible for the retention of serology typing and crossmatching as an adjunct to molecular HLA genotyping methodologies. In summary, genetic mutations can affect the level of expression for HLA and KIR and determine their isotype- and allotype-specific expression patterns. It is not always possible to predict those mutations from analysis of nucleotide sequence, and more expression variants may be uncovered in future functional analyses.

Genetic Polymorphism Affects Immune Function Amino acid substitutions can influence or change the function of either molecule or disrupt their interaction. Notwith-

Immunogenetics of HLA and KIR

69

Figure 3 KIR2DL and HLA-Cw genomic organization and interaction. Not to scale. KIR2DL1-3 and 3DL1-2 have similar genomic organization (nine exons; LP, leader peptide; D0-1, immunoglobulin domains; TM, transmembrane; Cyto, cytoplasmic). D0 is not transcribed in 2D KIR. Full white bar indicates conserved interaction site. Half white bar and dotted line indicate alternative interaction site. Residues 77 and 80 define Cw1 or 2; residue 77 (grey) not directly involved in contact with KIR. Dimorphisms at residue 80 of HLA-Cw and 44 and of KIR combine to alter binding site and specificity of interaction. Substitution at residue 68 of 2DL1 increases binding strength of 2DL1/Cw-2 compared with 2DL2/Cw-1.7

standing expression levels and tissue distribution, the basic functional difference conferred by HLA polymorphism is the peptide binding potential. A more extreme functional polymorphism is apparent for KIR: molecules capable of interacting with the same HLA class I may either inhibit or activate NK and T cells. These variants may be encoded by separate loci, such as 2DL1 and 2DS1, or alleles of the same locus, such as 3DL1/3DS1. Inhibitory KIR contain motifs in their cytoplasmic tails that are responsible for initiating the signaling cascade resulting in NK cell inhibition.9,10 Activating KIR have shorter cytoplasmic domains and associate with other molecules to transmit a stimulating signal.2 Most of the activating KIR occupy distinct loci, which are responsible for the majority of the KIR genotype diversity observed between human population groups (Fig 2). However, 3DS1 has probably arisen from an asymmetric recombination that replaced the inhibitory tail of 3DL1 with the tail from an activating KIR. The ligand for 3DS1 has not yet been identified but is generally assumed to be the same as for 3DL1, and they are alleles (Fig 1). Diversity correlates with the function of these molecules, and function may vary independently of HLA. However, for the receptor and ligand together, the importance of maintaining the integrity of their structural interaction becomes apparent. Further complexities to the HLA–KIR interaction are introduced by the peptide contained in the HLA molecule, the presence or absence of the first immunoglobulin domain of 3D KIR (D0), and by involvement of the LILR family of molecules.

Direct Evidence KIR interact with a motif of the HLA ␣-1 domain that is moderately conserved among HLA isotypes but differs suffi-

ciently to be distinguished both by KIR7 and alloreactive antibodies.13 HLA-Cw is split into two broad allele families, Cw1 and Cw2, by amino acid substitutions within this motif at positions 77 and 80. 2D KIR interact with HLA-Cw and a single substitution in the membrane-distal immunoglobulin domain (D1) defines their specificity for Cw1 or Cw2 (2DL1 interacts with Cw2, and 2DL2 with Cw1; Fig 3). Some 2D KIR (for example, 2DL2 and 2DS2) interact with identical Cw molecules, but a single amino acid change in D1 controls the strength of binding.6,10 In contrast, 2DS4, which is an activating KIR that is absent or severely mutated in many people, has a distinct D1 sequence that may explain its preference for only one HLA-Cw allotype (Cw04, which has a Cw2 motif) and its ability to bind an unknown ligand expressed by melanoma cell lines.32 That single amino acid substitutions change specificity lends credence to the proposal that the natural ligands of activating KIR may not be HLA. HLA-B alleles are also split into two broad families, termed Bw4 and Bw6, by substitutions within positions 77 to 83. Only HLA-B molecules containing the Bw4 motif are known to interact with KIR,1,6 in particular 3DL1 and possibly 3DS1. The Bw4 motif is further split by a substitution at position 80, which can be distinguished by the 3DL1 alleles. Some molecules have isoleucine at position 80 (such as at B51), which enables them to bind more strongly to 3DL1 than do those with threonine (such as B27) and better protects against NK cell–mediated cytolysis.33 However, the degree of NK cell control is also dependent on the 3DL1 allele. Differences in cytotoxic capability of NK cells could be attributed to single nucleotide substitutions in 3DL1 that change the strength of inhibition.33 Further complexities of 3DL1 functional diversity are revealed because these SNP sites were not the same as

P.J. Norman and P. Parham

70

Figure 4 HLA and KIR polymorphism. (Left) The ␣1 and ␣2 domains of HLA class I form the groove that binds antigenic peptide (circle) and exposes it to the TCR.13 The invariant ␤2-microglobulin (␤2M; not MHC-encoded) is noncovalently joined to ␣3. CD8 is a coreceptor on engagement of TCR complex (includes other molecules such as CD3; not shown). (Right) KIR binding resembles and overlaps TCR7; LILRB1 contacts ␣3 and ␤2M and may be the coreceptor in this interaction. Polymorphic residues that can affect interaction/inhibition are shown as stars. D2/␣2 interaction sites are conserved. Polymorphic sites may also influence KIR coaggregation during immune synapse formation.7

those controlling the different 3DL1 expression levels; neither were their positions likely to directly influence the KIR/ HLA interface.33 The two SNP sites that were investigated occurred in the transmembrane domain and at a site in D2 distal to the likely HLA interface (Fig 4). Crystal structures have not yet been obtained for the three immunoglobulin domain KIR/HLA heterodimers, but as most of the contact residues are the same, it is assumed that the interaction resembles that of the 2D KIR.7 The HLA- and peptide-contacting points occur near the junction between the two immunoglobulin domains of the KIR molecule. There are about 10 LILR isotypes, and two of the four with currently characterized ligands are also inhibitory receptors for HLA class I.19 LILRB1 interacts with most HLA class I molecules and is expressed by most immune effector cells.17,18 In antibody blocking experiments, KIR3DL1 and LILRB1 cooperated synergistically in their inhibition of NK cell–mediated cytolysis.33 The two SNP sites that increased the strength of 3DL1-mediated inhibition may then encourage coaggregation of the LILRB1 molecules33; equally likely is that the D0 domain encourages cooperation of KIR and LILRB1, because this domain has been shown to enhance binding of KIR to HLA.5 LILRB1 and KIR may behave in a similar fashion to CD8 and the TCR during their interaction with the HLA/peptide complex, so that LILRB1 acts as a coinhibitor. A model of this HLA/peptide/KIR/LILR interaction, illustrating all the suggested points of contact that are known to be polymorphic, is shown in Fig 4. LILR isotypes may also exhibit specificity for HLA allotypes,19 LILR can also be highly polymorphic,18 and their ligands are largely unidentified. The structural interaction of KIR3DL2 with HLA-A allo-

types A3 and A11 is peptide-specific.34 The only peptide fragment able to facilitate the interaction was derived from Epstein-Barr virus (EBV); HLA-A3 or A11 tetramers loaded with elf peptides did not interact with 3DL2, raising a new possibility for NK cell interactions. Are KIR-expressing cells “peptide-conditioned” during their maturation, or do some KIR act as germline-encoded pattern recognition molecules? That peptides derived from other viruses were not recognized by this particular KIR suggests the latter. On the other hand, as 3DL2 is an inhibitory molecule, this observation could indicate an elaborate evasion strategy by EBV, which is a successful parasite because it is resident for life in transformed B cells and rarely harms the host. The EBV nuclear protein (EBNA3A), from which the peptide is derived, is expressed during viral latency and may thus represent a state when the virus is effectively suppressing the immune response to cells that it has transformed. T-cell responses to latent EBV proteins tend to be highly focused and may be required to maintain normal virus life cycle.35 These memory T cells may be specifically inhibited by the 3DL2/peptide/ HLA interaction. Experiments investigating HLA-B27/NK cell binding have highlighted another interesting property of the HLA/KIR interaction that may be allele-dependent. Normally B27 interacts with 3DL1 because it possesses the Bw4 motif, but the HLA molecule also interacts with 3DL2. Under some conditions, B27 no longer binds to ␤2-microglobulin but exists on the cell surface as a homodimer that retains some peptide binding capability.19 This heavy-chain homodimer gains ability to interact with 3DL2 but loses ability to interact with LILRB1, implying that the NK or memory T-cell repertoire may be able to distinguish the two states.19 Although homodimerization is not unique to B27 or disease, HLA-B*27 has long been recognized as a risk factor for spondyloarthropathy; the unusual mechanics of the molecular interaction warrant further investigation of in this poorly understood disease.19 In summary, naturally occurring amino acid substitutions in either HLA or KIR molecules can abrogate or enhance in vitro functional capability of immune effector cells, due to direct influences at the structural interface or indirectly via mechanisms or molecules that have not yet been elucidated. Further evidence that genetic variation in either of these molecules can alter immune activity arises from genetic associations that have implicated both. Other than transplantation, KIR/HLA function may be explored also in infectious diseases and pregnancy, while the most obvious dysfunction should become apparent from the collateral damage caused by autoimmunity.

Genetic Associations Linkage Disequilibrium Success in HLA disease association studies derives in part from the close relationship between genetic variations and co-segregation of their loci. Linkage disequilibrium (LD) occurs when alleles of two (or more) genes encoded on the same chromosome occur together more frequently than would be

Immunogenetics of HLA and KIR expected by random association and results in discernible haplotypes in multiple unrelated individuals. LD is an essential tool for immunogeneticists, but the reasons for association are many and not always distinct.36 Both the MHC and KIR exhibit examples of strong LD that are not always consistent with the physical distance between their respective loci.37,38 DNA recombination during gametogenesis disrupts LD, and recombination may occur more frequently in some genetic regions, leading to the formation of localized recombination “hotspots” and LD blocks, phenomena that were originally identified and have been principally characterized in the human MHC.14,39 Some consistency is observed between ethnic populations; blocks and hotspots along the genetic region encompassing HLA-B and -Cw, for example, can occur in the same place in distinct population groups, despite differing allele compositions and frequencies.14 Ahmad et al present an excellent visual interpretation of LD and its relationship to MHC haplotypes.37 As the KIR and MHC are both multigene families that exhibit complicated patterns of blocks and hotspots,21,37 it may be difficult to identify the causative agent once a statistical disease association has been found. For example, psoriasis was originally associated with a HLA-Cw allele (Cw*06), but this link was not always reproducible between population groups, suggesting more than one disease lesion or a causal gene in LD with the offending HLA. Both explanations now seem likely, as haplotypes that contain Cw*06 and the SNP that causes upregulation of corniodesmosin are associated with this disease.24 The corniodesmosin gene is 100 kb from HLA-Cw and is abnormally expressed by keratinocytes during psoriasis. Although environment and chance dictate that we are not all subject to the worst infectious diseases, we have all assumed the position of fetus at least once. Maternal NK cells are the predominant immune cell type in the interface between mother and fetus during early development. Extravillous trophoblast cells (EVT) invade the uterus, destroy arterial walls, and replace endothelial cells, thus ensuring the fetus with an adequate blood supply.28 Uterine NK cells probably control this process, such that poor supervision of EVT leads to poor vascularization and the perilous condition of preeclampsia.28 EVT express a limited array of HLA molecules that distinguishes them from other fetal cells and the mother’s cells. Uterine NK cells in turn are distinguished from circulating NK cells.6 Possession of HLA-Cw2 by the fetus accompanied by 2DL1 in the mother may predispose towards preeclampsia; the chance of disease development then decreases as the number of activating receptors increases.40 Thus, mothers who are homozygous for the “A” haplotype (and have few activating KIR; see Fig 1) may be more susceptible to preeclampsia if confronted with a Cw2-bearing embryo, a factor that has led over time to HLA-Cw2 and KIR A haplotype having reciprocal frequencies in human populations.40 Preeclampsia may arise because 2DL1 is more discriminating than 2DL2/3 for HLA-Cw allotype recognition and has stronger binding kinetics (Fig 3). The rapid association/disassociation rates for KIR-HLA interactions enable NK cells to respond to changes in HLA expression levels.6,7 Stronger binding of Cw2 and 2DL1 could mean that this

71 interaction is less sensitive to the subtle changes in HLA expression that may be necessary for successful EVT invasion. In addition to controlling implantation, uterine NK cells may also regulate the aggressive phase of the menstrual cycle, when the uterine wall is broken down.28 Increased 2DL1 expression has been observed in patients with endometriosis, in which lack of control over breakdown and inappropriate growth of uterine tissue may ultimately interfere with implantation.41 KIRs 2DL2 and 3 are alleles and their products perform the same function.5 However, their differing affinities for HLACw1 lead to a similar disease model for hepatitis C to that described for preeclampsia.42 As 2DL3 is a less potent inhibitor and has lower affinity for Cw1 than does 2DL2,7,10 individuals who are homozygous for both Cw1 and 2DL3 are able to clear low-dose infections more efficiently than does any other genotype combination.42 NK cells in 2DL3 homozygous individuals are more easily activated when they have only Cw1 to inhibit them. The activating receptors that respond to hepatitis C are unlikely to be KIR because 2DL3 homozygotes do not have many activating KIR (Fig 1). When Cw2 is present it is recognized by 2DL1 (because 2DL1 is virtually universal [Fig 1]) and the cells that would respond to hepatitis C are harder to activate. Furthermore, 2DL2 counteracts the advantage incurred by 2DL3 because it binds more strongly to the HLA molecule, and inhibitory KIR can override activation by most other receptors.42 A Case of Missing Identity NK cells in transplantation settings may be concerned with missing self,3 while those encountering virus infection cannot be blamed for their mistaken identity of self impersonators.2 Unwelcome immune activity, on the other hand, may arise when the “self” was never there. Psoriatic arthritis may result from reckless activating KIR but is more likely to occur if inhibitory KIR are unable to control NK cell activity because their HLA ligand is not present.43 KIR “B” haplotypes produce a higher ratio of NK cells that express activating receptors (Fig 1). Any excess activity should be moderated when inhibitory receptors recognize HLA. In HLA-Cw1 or -2 homozygous individuals who also possess predominantly KIR B haplotypes, this control is weakened and so the missing HLA molecule may predispose to psoriatic arthritis. Highlighting the complexity of such disease associations, it was initially unclear whether the most influential contributing factor was lack of an inhibitory, or the increase of an activating response23; their balance now appears most important.43 In summary, genetically defined differences in HLA and KIR that determine the strength of their interaction correlate with resistance or susceptibility to disease. Like many disease associations, the biologic mechanisms remain to be proven, but corroborative evidence is accumulating.

Polymorphism and Evolution Evolution and structure/function are interdependent, such that the presence of a multitude of genetic variations, each with its own impact on host response, provides good evi-

72 dence for creation of a constantly adapting innate immune system. The high variety observed of HLA and KIR indicates that positive selection has encouraged mutations to be retained. Selection events may leave a distinct, albeit fading, impression on genetic architecture that may be exposed during population and genomic studies. Repeat mutation is unlikely (although not impossible) so that once a new allele arises, haplotype diversification is only achieved by recombination and/or coincident mutation proportional with time. Thus a more divergent haplotype displaying weak LD with flanking markers suggests an older mutation. Conversely, an allele displaying strong LD with flanking markers and low diversity implies relatively newer polymorphism and/or the increase in frequency of a certain haplotype due to natural selection (although a similar effect could be caused by population admixture or historical bottleneck and genetic drift combined with massive population expansion over the last millennium45). These two extremes can be illustrated by the HFE gene, which is located in the telomeric extremity of the MHC and has two mutations that associate with hemochromatososis. One of these (H06D) must be relatively old as it is dispersed among different haplotypes and human population groups. The other mutation (C282Y) was identified by association with a common HLA haplotype that encodes A1 and B8. Selection, as revealed by low diversity37 and long-range LD of this haplotype, has likely favored the HFE mutation rather than one of the HLA alleles.44 The once appealing hypothesis that the Y282 mutation arose in the last millennium and hitch-hiked around Western Europe with marauding Vikings is being supplanted by suggestions that the mutation arose during Neolithic times when our ancestors were undergoing dietary and environmental changes.44 The potential for high iron uptake has obvious benefits and may present a better rationale than infectious disease-resistance to explain the prevalence of the A1-B8 haplotype, which is also associated with several autoimmune diseases in European Caucasians. The imprint of natural selection remains on KIR haplotypes also. KIR3DS1 is present in approximately 25% of Caucasians and is almost exclusively carried by a single SNP haplotype, suggesting that 3DS1 or another closely linked KIR/allele has reached high frequency due to positive selection.38 The nonexpressed 3DL1*004 represents 20% of 3DL1 alleles in the Caucasian population46 and is also associated with a distinct KIR SNP haplotype,38 suggesting it too has been naturally selected. In another example, the recombination that has enabled 3DP1 expression by some haplotypes may be recent.29 These activated 3DP1 genes were predominantly associated with identical duplications of the remainder of the central framework genes, 2DL4 and 3DL1. The KIR A haplotype appears much more diversified, but this may simply reflect its higher frequency; A haplotypes have a varied set of alleles and SNP markers despite their consistent gene number.5,38 As new genetic variants arise and expand, others may diminish, as has been inferred from many population-based HLA studies.14,47 Alleles or genes may be lost by natural wastage (genetic drift) or if they become a liability. There may

P.J. Norman and P. Parham have been pressure to actively disable 3DL1 in some populations, resulting in the high frequency of 3DL1*004. Should a virus evolve to evade NK cells by molecular mimic of HLABw4, it would become a disadvantage to express 3DL1 and also possibly an advantage to possess 3DS1, similar to the situation in mice.2 3DL1*004 could have increased in frequency by genetic drift; however, if we include all of the alleles of this locus and consider 3DS1 simply as a lack of 3DL1, then less than half of chromosomes carry a functionally “normal” 3DL1 in Caucasians at least.38,46

Heterozygosity and Balancing Selection HLA molecules have been under positive balancing selection, which ensures variety and enhances heterozygosity by maintaining allele groups at relatively equivalent frequencies within populations.16,47 This selection pressure has also resulted in distinct subsets of HLA alleles for the various human population groups47 (see also www.allelefrequencies.net). Balancing selection maintains LD between loci in addition to their diversity,48 such that finding common selection-enhanced alleles can be justified and ancient haplotypes predating modern diversification of humans can be distinguished from a highly heterogeneous background.13,16 The distributions of KIR haplotypes between and 3DL1 alleles within human populations demonstrate that KIR have been subject to balancing selection.5,22,23,38,46 Two alternative but not mutually exclusive models of balancing selection are heterozygous advantage and frequencydependent selection. Studies of human immunodeficiency virus (HIV) exemplify both. Heterozygous advantage is apparent when HLA heterozygous individuals can respond to a wider range of HIV antigens.49 Frequency-dependent selection, which favors rare HLA alleles over time, is illustrated when HIV escape mutants accumulate in successively infected individuals, so that those having common HLA types may be exposed to “pre-evolved” virus. Conversely, selection pressure may be exerted on viral development by HLA.49 Some HLA alleles (B*57 and *5801, which are relatively rare) are associated with slow progression to AIDS because they can limit the generation of viable viral mutants, so that those mutants able to escape may be replication-compromised. In these cases, the mutants revert to wild type when they are freed of these constraints by secondary transmission into individuals with different HLA alleles.49 Resistance to hepatitis C may also illustrate the frequency-dependent model of balanced selection, in that 2DL3/Cw1 homozygous individuals who have an advantage specifically related to this infection may not be equipped for the next common infectious agent.42 KIR have also been implicated in HIV: infected individuals who possess 3DS1 and an appropriate Bw4 motif have slower disease progression.23 Activating 3DS1 might promote lysis of infected cells, possibly by responding to foreign peptide. Individuals who are homozygous for 2DL3 generally do not have a copy of 3DS1 (Fig 1), and so hepatitis C–resistant individuals may not also be capable of slowing HIV progression. It seems likely from disease association studies that the KIR

Immunogenetics of HLA and KIR haplotype represents the functional genetic unit. Possession of a diverse array of activating receptors may increase resistance to particular diseases and thus confer a selection advantage, or KIR B haplotypes may simply result in a higher proportion of NK cells expressing activating receptors, so that innate immunity is more easily triggered. In contrast, the A haplotypes encode few or no21 activating receptors, so that natural immune responses may be more tempered. The prevalence of such functionally opposed haplotypes in the human population is probably the result of a balancing selection, or a trade-off between insufficient reaction and over-reaction leading to autoimmunity. We can now bolster this concept to include detrimental inactivity during pregnancy, and, tentatively suggest that classic examples of immune over-reaction such as lepromatous (cytotoxic) leprosy, visceral leishmaniasis, and cerebral malaria may maintain the balance.

KIR Today KIR genes and haplotypes have a complex ancestry that clearly demonstrates their rapid evolution under strong natural selection pressure20; to be ascertained is whether KIR variation has arisen in response to the constantly evolving HLA or other possibly extrinsic factors. The Caribbean purple-throated hummingbird Eulampis jugularis has a polymorphic bill that enables it to feed from a polymorphic flower, of which it is the sole pollinator. On different islands the beaks and the flowers have diverged in size and structure but retained complimentary morphologies and interdependence. Their continued interaction allows the bird to feed while the flower is still pollinated, demonstrating their reciprocal evolution.50 To determine whether KIR and HLA structural conformity has been so elegantly synchronized will require first understanding the true extent of KIR polymorphism in contemporary human populations and the relationship with HLA allele distribution. Some HLA alleles have arisen in the last few centuries,51 confirming that HLA evolution is an ongoing process, but it is unknown whether KIR are still evolving rapidly. The diversity of KIR locus haplotypes among populations suggests that their positive evolution is current,5 whereas the propensity for locus and allele inactivation (pseudogenes, 2DL5B, 2DS4del, 3DL1*004, 2DL4-9a) implies what was once required might no longer be an asset. Perhaps KIR were needed at one point in our evolutionary history to combat particular diseases and later became a liability. The structural constraints placed on KIR by HLA polymorphism could have led to this “boom and bust” evolution. HLA molecules must retain the ability to bind varying peptide and to the TCR, while protecting cells from NK cell–mediated death. Furthermore, the huge interspecies diversity of NK cell receptors parallels the difference in implantation strategies among mammals,40 so that more aggressive invasion requires a more carefully controlled response. Activating KIR that recognize pathogen or tumor factors must not mediate overzealous killing of healthy cells. Compounded by population amalgamation, migration and expansion, and the development of modern

73 medicine and societies, the frontier of KIR evolution may have been caught between many factions. The molecular mechanism of ligand/receptor recognition may help us to understand its evolution. That KIR-contacting residues inhabit a region of HLA of otherwise highly variable nature does indicate evolutionary constraints placed on the interaction, although it is not clear whether KIR have evolved to interact with this region or the reverse. In addition to the D1/␣-1 interaction that determines specificity, KIR known to bind HLA have well conserved sequences in D2 (Fig 3). These residues interact with a motif in the ␣-2 domain, which is almost identical in their HLA-B and -Cw ligands.7 D2/␣-2 may then provide the linchpin that has enabled the D1/␣-1 interaction to develop gradually over time. As the receptors have developed, new properties have emerged. For example, the increased strength of 2DL1 binding caused by a single SNP (Fig 3) may have led to the downfall of its Cw2 ligand in some populations.40 Moreover, 11 of the 12 HLA-Cw residues that contact KIR are invariant and the only difference occurs with position 807; residue 80 is further mutated in HLA Bw4 and defines 3DL1 specificity for this motif. Have position 80 of HLA and the corresponding amino acid of KIR molecules been co-evolving, or have KIR just turned up late and tried to fit in?

Conclusion The most important impact of HLA variation on transplantation remains generation of antibody- or TCR-mediated responses towards HLA molecules that are recognized as foreign. Nevertheless, KIR compatibility enforced upon transplanted tissue or cells is becoming increasingly relevant with improvement of grafting techniques and reduction of immunosuppression. Even at low resolution, less than 1% KIR genotypes will be repeated in unrelated individuals.5,38 Together with variation between populations, this diversity alerts us to the need to define adequate genotyping techniques for use in transplantation in order to encompass all ethnic groups. Like HLA, finding a perfect KIR match in unrelated individuals will be a challenge, should it be required. Finding a perfect KIR and HLA match would be difficult, even in related individuals.5 Fortunate, then, the cornerstone of clinical transplantation studies has been the clear demonstration that an NK cell–mediated response can remove leukemic cells and prevent graft-versus-host disease when KIR and HLA are suitably mismatched.3 KIR genes like HLA genes have many alleles, but gene content variation was the first level of their diversity to be recognized.5 Most of the analyses described to date will have been undertaken at this early stage. Without consideration for differences in functional characteristics of the different alleles, the immunologic relevance of a statistical association can never be fully explained. Furthermore, when viewed in the wider perspective that often (especially in the case of transplantation) includes variation in treatment regimens, we should not be too surprised to encounter apparent contradictions. These studies have alerted us to the importance in many clinical settings of HLA and KIR, and most importantly, their interaction. Although well explored for HLA,

P.J. Norman and P. Parham

74 we are still fathoming the extent and nature of KIR genes and alleles, but we are beginning to appreciate the substantial impact of their genetically determined variation upon our individual variations in the immune response to infection, malignancy, and transplantation.

22.

23.

Acknowledgment

24.

To those we could not reference due to space and to Michael Gleimer, Linh Pham, and Laurent Abi-Rached for suggestions, many thanks.

25.

26.

References 1. Moretta L, Moretta A. Unravelling natural killer cell function: triggering and inhibitory human NK receptors. EMBO J 23:255-259, 2004 2. Lanier LL: NK cell recognition. Annu Rev Immunol 23 (doi: 10.1146/ annurev.immunol.23.021704.115526), 2005 3. Ruggeri L, et al: Role of natural killer cell alloreactivity in HLA-mismatched hematopoietic stem cell transplantation. Blood 94:333-339, 1999 4. Trowsdale J, Parham P: Mini-review: Defense strategies and immunityrelated genes. Eur J Immunol 34:7-17, 2004 5. Vilches C, Parham P: KIR: Diverse, rapidly evolving receptors of innate and adaptive immunity. Annu Rev Immunol 20: 217-251, 2002 6. Vales-Gomez M, Reyburn H, Strominger J: Interaction between the human NK receptors and their ligands. Crit Rev Immunol 20:223-244, 2000 7. Boyington JC, Sun PD: A structural perspective on MHC class I recognition by killer cell immunoglobulin-like receptors. Mol Immunol 38: 1007-1021, 2002 8. Biron CA, Nguyen KB, Pien GC, Cousens LP, Salazar-Mather TP: Natural killer cells in antiviral defense: Function and regulation by innate cytokines. Annu Rev Immunol 17: 189-220, 1999 9. Long EO: Tumor cell recognition by natural killer cells. Semin Cancer Biol 12:57-61, 2002 10. Campbell KS, Colonna M: Human natural killer cell receptors and signal transduction. Int Rev Immunol 20:333-370, 2001 11. Zingoni A, Sornasse T, Cocks BG, Tanaka Y, Santoni A, Lanier LL: Cross-talk between activated human NK cells and CD4⫹ T cells via OX40-OX40 ligand interactions. J Immunol 173:3716-3724, 2004 12. Raulet DH: Interplay of natural killer cells and their receptors with the adaptive immune response. Nat Immunol 5:996-1002, 2004 13. Marsh SG, Parham P, Barber LD: The HLA Facts Book. FactsBook. London, UK: Academic Press, 2000 14. Jeffreys AJ, Holloway JK, Kauppi L, May CA, Neumann R, Slingsby MT, et al: Meiotic recombination hot spots and human DNA diversity. Philos Trans R Soc Lond B Biol Sci 359:141-152, 2004 15. Messaoudi I, Guevara Patino JA, Dyall R, LeMaoult J, Nikolich-Zugich J: Direct link between mhc polymorphism, T cell avidity, and diversity in immune defense. Science 298:1797-1800, 2002 16. Hughes AL, Yeager M: Natural selection at major histocompatibility complex loci of vertebrates. Annu Rev Genet 32: 415-435, 1998 17. Cosman D, Fanger N, Borges L, Kubin M, Chin W, Peterson L, et al: A novel immunoglobulin superfamily receptor for cellular and viral MHC class I molecules. Immunity 7:273-282, 1997 18. Colonna M, Nakajima H, Navarro F, Lopez-Botet M: A novel family of Ig-like receptors for HLA class I molecules that modulate function of lymphoid and myeloid cells. J Leukoc Biol 66:375-381, 1999 19. Allen RL, Trowsdale J: Recognition of classical and heavy chain forms of HLA-B27 by leukocyte receptors. Curr Mol Med 4:59-65, 2004 20. Rajalingam R, Parham P, Abi-Rached L: Domain shuffling has been the main mechanism forming new hominoid killer cell Ig-like receptors. J Immunol 172:356-369, 2004 21. Hsu KC, Chida S, Dupont B, Geraghty DE: The killer cell immunoglob-

27.

28. 29.

30.

31.

32.

33. 34. 35. 36. 37.

38.

39.

40.

41.

42.

ulin-like receptor (KIR) genomic region: Gene-order, haplotypes and allelic polymorphism. Immunol Rev 190: 40-52, 2002 Yawata M, Yawata N, Abi-Rached L, Parham P: Variation within the human killer cell immunoglobulin-like receptor (KIR) gene family. Crit Rev Immunol 22:463-482, 2002 Carrington M, Norman P: The KIR Gene Cluster. Bethesda, MD: National Library of Medicine, National Center for Biotechnology Information, 2003 Capon F, Trembath RC, Barker JN: An update on the genetics of psoriasis. Dermatol Clin 22:339-347, 2004 Trompeter HI, Gomez-Lozano N, Santourlidis S, Eisermann B, Wernet P, Vilches C, et al: Three structurally and functionally divergent kinds of promoters regulate expression of clonally distributed KIR, of KIR2DL4, and of KIR3DL3. J Immunol (in press) Uhrberg M: Shaping the human NK cell repertoire: An epigenetic glance at KIR gene regulation. Mol Immunol 42:471-475, 2004 Goodridge JP, Witt CS, Christiansen FT, Warren HS: KIR2DL4 (CD158d) genotype influences expression and function in NK cells. J Immunol 171:1768-1774, 2003 Moffett-King A: Natural killer cells and pregnancy. Nat Rev Immunol 2:656-663, 2002 Gomez-Lozano N, Estefania E, Williams F, Halfpenny IA, Middleton D, Solis R, et al: The silent KIR3DP1 gene (CD158c) is transcribed and might encode a secreted receptor in a minority of humans, in whom the KIR3DP1, KIR2DL4 and KIR3DL1/KIR3DS1 genes are duplicated. Eur J Immunol 35:16-24, 2005 Pando MJ, Gardiner CM, Gleimer M, McQueen KL, Parham P: The protein made from a common allele of KIR3DL1 (3DL1*004) is poorly expressed at cell surfaces due to substitution at positions 86 in Ig domain 0 and 182 in Ig domain 1. J Immunol 171:6640-6649, 2003 Norman PJ, Carrington CV, Byng M, Maxwell LD, Curran MD, Stephens HA, et al: Natural killer cell immunoglobulin-like receptor (KIR) locus profiles in African and South Asian populations. Genes Immun 3:86-95, 2002 Katz G, Gazit R, Arnon TI, Gonen-Gross T, Tarcic G, Markel G, et al: MHC class I-independent recognition of NK-activating receptor KIR2DS4. J Immunol 173:1819-1825, 2004 Carr WH, Pando MJ, Parham P: KIR3DL1 polymorphisms that affect NK cell inhibition by HLA-Bw4 ligand. (manuscript in preparation) Hansasuta P, et al: Recognition of HLA-A3 and HLA-A11 by KIR3DL2 is peptide-specific. Eur J Immunol 34:1673-1679, 2004 Rickinson AB, Moss DJ: Human cytotoxic T lymphocyte responses to Epstein-Barr virus infection. Annu Rev Immunol 15: 405-431, 1997 Pritchard JK, Przeworski M: Linkage disequilibrium in humans: models and data. Am J Hum Genet 69:1-14, 2001 Ahmad T, Neville M, Marshall SE, Armuzzi A, Mulcahy-Hawes K, Crawshaw J, et al: Haplotype-specific linkage disequilibrium patterns define the genetic topography of the human MHC. Hum Mol Genet 12:647-656, 2003 Norman PJ, Cook MA, Carey BS, Carrington CV, Verity DH, Hameed K, et al: SNP haplotypes and allele frequencies show evidence for disruptive and balancing selection in the human leukocyte receptor complex. Immunogenetics 56:225-237, 2004 Alper CA, Raum D, Karp S, Awdeh ZL, Yunis EJ: Serum complement ’supergenes’ of the major histocompatibility complex in man (complotypes). Vox Sang 45:62-67, 1983 Hiby SE, Walker JJ, O’Shaughnessy KM, Redman CW, Carrington M, Trowsdale J, et al: Combinations of maternal KIR and fetal HLA-C genes influence the risk of preeclampsia and reproductive success. J Exp Med 200:957-965, 2004 Maeda N, Izumiya C, Yamamoto Y, Oguri H, Kusume T, Fukaya T: Increased killer inhibitory receptor KIR2DL1 expression among natural killer cells in women with pelvic endometriosis. Fertil Steril 77:297302, 2002 Khakoo SI, Thio CL, Martin MP, Brooks CR, Gao X, Astemborski J, et al: HLA and NK cell inhibitory receptor genes in resolving hepatitis C virus infection. Science 305:872-874, 2004

Immunogenetics of HLA and KIR 43. Nelson GW, Martin MP, Gladman D, Wade J, Trowsdale J, Carrington M: Cutting edge: heterozygote advantage in autoimmune disease: Hierarchy of protection/susceptibility conferred by HLA and killer Ig-like receptor combinations in psoriatic arthritis. J Immunol 173:4273-4276, 2004 44. Distante S, Robson KJ, Graham-Campbell J, Arnaiz-Villena A, Brissot P, Worwood M: The origin and spread of the HFE-C282Y haemochromatosis mutation. Hum Genet 2004 45. Bamshad M, Wooding SP: Signatures of natural selection in the human genome. Nat Rev Genet 4:99-111, 2003 46. Halfpenny IA, Middleton D, Barnett YA, Williams F: Investigation of killer cell immunoglobulin-like receptor gene diversity: IV. KIR3DL1/ S1. Hum Immunol 65:602-612, 2004

75 47. Cavalli-Sforza L, Menozzi P, Piazza A: The History and Geography of Human Genes. Princeton, NJ: Princeton University Press, 1994 48. Slatkin M: Balancing selection at closely linked, overdominant loci in a finite population. Genetics 154:1367-1378, 2000 49. Altman JD, Feinberg MB: HIV escape: There and back again. Nat Med 10:229-230, 2004 50. Temeles EJ, Kress WJ: Adaptation in a plant-hummingbird association. Science 300:630-633, 2003 51. Trachtenberg EA, Erlich HA, Rickards O, DeStefano GF, Klitz W: HLA class II linkage disequilibrium and haplotype evolution in the Cayapa Indians of Ecuador. Am J Hum Genet 57:415-424, 1995