Missing or altered self: human NK cell receptors that recognize HLA-C

Natural killer (NK) cells are lymphocytes that are critical to the early phase of an immune response, for antibody-dependent cytotoxicity (ADCC) and development of the placenta in early pregnancy ( Parham and Moffett 2013 ). The response of NK cells to infected, cancerous or semi-allogeneic tissue is guided by interactions between activating or inhibitory NK cell receptors and major histocompatibility complex (MHC) class I ligands (known as HLA class I in humans). These direct interactions enable NK cells to distinguish healthy cells from diseased cells, which commonly have altered expression or abnormal forms of MHC class I. In mammalian species, NK cell receptors typically comprise a bipartite system. This combines conserved receptors that recognize non-polymorphic MHC class I with diverse receptors that recognize polymorphic MHC class I ( Guethlein et al. 2015 ). In humans, HLA-E binds a restricted set of peptides, which are largely derived from the leader sequences of the polymorphic HLA class I molecules, and thus forms the ligand for CD94:NKG2 lectin-like receptors ( Braud et al. 1998 ; Horowitz et al. 2016 ). Because both receptor and ligand are genetically conserved, these interactions are a constant feature of human immune responses. In contrast, polymorphic HLA-A, B and C bind a diverse repertoire of peptides and are recognized by the killer-cell immunoglobulin-like receptors (KIR). Because both HLA-A, B and –C and KIR are highly polymorphic, their interactions diversify and individualize NK cell responses ( Valiante et al. 1997 ). The functional importance of these diverse interactions is illustrated by the association of various combinations of KIR and HLA class I with the outcome of infection ( Alter et al. 2011 ; Holzemer et al. 2015 ; Khakoo et al. 2004 ; Wauquier et al. 2010 ), susceptibility to autoimmune disease ( Luszczek et al. 2004 ; Suzuki et al. 2004 ; van der Slik et al. 2003 ) and reproductive success ( Hiby et al. 2004 ; Hiby et al. 2010 ; Nakimuli et al. 2015 ). KIR emerged comparatively recently in evolutionary history and are largely restricted to simian primates (monkeys, apes and humans) ( Guethlein et al. 2015 ). The KIR gene family is diversified by both gene content variability and allelic polymorphism. However, the basic organization of the KIR locus is conserved amongst the catarrhine primates (Old World monkeys and apes) ( Parham et al. 2010 ). It consists of four framework genes, which are common to all haplotypes, and a suite of homologous genes that are variably present among species and individuals ( Wilson et al. 2000 ). A recombination site at the center of the locus defines two distinct regions: one closer to the chromosomal centromere (termed centromeric), the other closer to the chromosomal telomere (termed telomeric) ( Pyo et al. 2013 ). Humans KIR haplotypes uniquely form two distinctive groups: KIR A haplotypes encode a fixed suite of largely inhibitory KIR whereas KIR B haplotypes have a variable number of inhibitory receptors and several activating receptors ( Uhrberg et al. 1997 ). KIR bind to the upper face of the HLA class I molecule over the C-terminal end of the bound peptide ( Boyington et al. 2000 ; Fan et al. 2001 ; Liu et al. 2014 ). Key polymorphisms within the α 1 helix of the HLA molecule determine four major epitopes (A3/11, Bw4, C1 and C2) recognized by KIR ( Saunders et al. 2015 ). Whereas a minority of HLA-A and HLA-B encode the A3/11 and Bw4 epitopes, all HLA-C have either the C1 or C2 epitope. Consequently, it is the interactions between HLA-C and the two-domain inhibitory and activating KIR that recognize HLA-C and dominate human NK cell responses ( Colonna et al. 1993 ; Valiante et al. 1997 ). The two mutually exclusive HLA-C epitopes are defined by dimorphism at positon 80 in the α 1 helix. The HLA-C1 epitope is defined by N80 and recognized by KIR with K44 (KIR2DL2/3, K44-KIR2DP1 F , KIR2DS2), the HLA-C2 epitope is defined by K80 and recognized by KIR with either M44 (KIR2DL1, KIR2DS1) or T44 (KIR2DS3/5, T44-KIR2DP1 F ) ( Hilton et al. 2017b ; Moesta et al. 2008 ; Moesta et al. 2010 ; Winter et al. 1998 ). Here we review the evolution and function of human NK cell receptors that recognize HLA-C, detailing the functional consequences of their extreme diversity and the mechanisms by which they discriminate healthy cells from diseased and semi-allogeneic cells. Human inhibitory KIR recognize four mutually exclusive epitopes on HLA class I. The KIR that recognize these epitopes comprise two lineages that have a common genetic organization amongst the catarrhine primates (Old World monkeys and apes) ( Guethlein et al. 2007 ; Wilson et al. 2000 ). Lineage II KIR recognize the A3/11 and Bw4 epitopes of HLA-A and HLA-B whereas the lineage III KIR recognize the C1 and C2 epitopes of HLA-C ( Moesta et al. 2008 ). Comparative studies examining the organization of the KIR and HLA class I loci in non-human primates have provided insight into the co-evolution of these immune system elements and contributed to our understanding of their functions in reproduction and immunity in modern humans. Old world monkeys, such as the rhesus macaque, have counterparts to HLA-A and HLA-B , but not to HLA-C ( Adams and Parham 2001 ). In this species, an abundance of MHC-A and- B genes that encode the Bw4 epitope is accompanied by a corresponding expansion of the lineage II KIR that recognize them ( Bimber et al. 2008 ; Blokhuis et al. 2010 ; Kruse et al. 2010 ). As a result, the telomeric region of the rhesus macaque KIR locus contains 21 lineage II KIR genes. By contrast, the centromeric region of the rhesus macaque KIR locus contains just one lineage III KIR , and its function is not known ( Bimber and Evans 2015 ). The organization of the orangutan and chimpanzee KIR loci is the reverse. Corresponding with the emergence and fixation of MHC-C , the centromeric region of the KIR locus contains different combinations of nine lineage III KIR genes encoding receptors that recognize the C1 or C2 epitopes ( Abi-Rached et al. 2010 ; Older Aguilar et al. 2010 ). The telomeric region has just one lineage II KIR encoding a receptor for Bw4-like epitopes of MHC-A and MHC-B. In humans, the A3/11 and Bw4 epitopes are carried by a minority subset of HLA-A and HLA-B allotypes. By contrast, all HLA-C allotypes have either the C1 or C2 epitope ( Parham and Moffett 2013 ). The result is that only ~50% of the human population have Bw4 and/or A3/11, whereas 100% have C1 and/or C2. By this simple measure, HLA-C is seen to be the dominant source of ligands for human inhibitory KIR. Together, these studies are consistent with MHC-C having arisen under natural selection in the higher primates to be a more specialized ligand for KIR than either HLA-A or HLA-B ( Older Aguilar et al. 2010 ). Contributing to the qualitative and quantitative changes in the hominid KIR locus was an increasing role in reproduction for interactions between MHC-C and KIR , specifically in the formation of the placenta ( Moffett and Colucci 2015 ). Whereas HLA-A, -B and –C are co-expressed by most cells of the body, a crucial difference occurs in pregnancy, where extravillous trophoblast cells express HLA-C but not HLA-A or HLA-B ( King et al. 2000 ). Extravillous trophoblast are cells of fetal origin that invade the mother’s uterus, where they transform the spiral arteries into large vessels, coopting sufficient maternal blood to sustain the developing fetus until the end of pregnancy ( Moffett-King 2002 ). This invasive placental phenotype observed in hominids is absent in Old World monkeys, and thus correlates with the emergence of MHC-C in the orangutan. Comparing the KIR of different primate species shows an extraordinary degree of species-specificity that attests to their rapid evolution. This specificity is particularly evident from comparison of the human and chimpanzee lineage III KIR . During the speciation event that separated humans from chimpanzees ~5 million years ago, the chimpanzee line retained much of the lineage III KIR diversity that had evolved in the previous 10–15 million years and was present in the last common ancestor. By contrast, owing to population bottlenecks during speciation, humans retained just two lineage III KIR : one encoding activating KIR2DS4 and the other encoding an inhibitory MHC-C receptor ( Abi-Rached et al. 2010 ; Graef et al. 2009 ). The latter gene subsequently became subject to a process of gene duplication and differentiation that spawned a family of six human-specific KIR genes and formation of the distinctive CenA and CenB KIR haplotypes. Our recent study suggests this founder gene was KIR2DP1 F , the functional antecedent of the now-inactive KIR2DP1 gene ( Hilton et al. 2017a ). KIR2DP1 F is likely to have first been a dedicated and polymorphic C1-specific receptor with K44 as the specificity determining residue ( Figure 1 ). Subsequently, it evolved a second lineage of alleles with T44 and C2 specificity ( Hilton et al. 2017a ). In this way, T44 KIR2DP1 F served to replace the M44 lineage III KIR with C2-specificity that were lost from the human lineage during speciation ( Abi-Rached et al. 2010 ; Hilton et al. 2017a ). Provision of a single receptor with alternative allotypes that could recognize either the C1 or C2 epitopes was a major step forward. However, although useful as an evolutionary intermediate, KIR2DP1 F was subsequently usurped by K44 KIR2DL2/3 and M44 KIR2DL1, dedicated high affinity receptors with specificity for the C1 and C2 epitopes of HLA-C, respectively. KIR2DP1 F became attenuated and subsequently inactivated due to the acquisition and propagation of a single nucleotide deletion that caused premature termination of the protein ( Hilton et al. 2017a ; Vilches et al. 2000 ). Inactivation of KIR2DP1 F was complete prior to the divergence of modern humans from archaic humans some 450,000 years ago ( Figure 1 ) ( Hilton et al. 2017a ). An early feature of human-specific evolution was the establishment of two functionally distinctive KIR haplotypes ( Uhrberg et al. 1997 ). CenA haplotypes are characterized by strong and specific inhibitory KIR allotypes, whereas CenB haplotypes have weakened inhibitory KIR allotypes and a variable number of activating KIR ( Hilton et al. 2015a ). KIR2DP1 F allotypes contributed a key first step in this evolution. Higher affinity K44 KIR2DP1 F allotypes accumulated in the centromeric part of CenA whereas lower affinity T44 KIR2DP1 F allotypes, many of which also had attenuated or inactivated signaling function, accumulated in CenB ( Hilton et al. 2017a ). KIR2DL1, which is in strong linkage disequilibrium with KIR2DP1 shows a similar evolutionary trajectory. Strong KIR2DL1 allotypes with high affinity for HLA-C2 and intact signaling function segregate on CenA , whereas weaker KIR2DL1 allotypes with reduced affinity for HLA-C2 and/or attenuated signaling function are found on CenB ( Bari et al. 2009 ; Hilton et al. 2015a ). The trend for inhibitory KIR with reduced capacity to recognize or respond to HLA-C2 is taken to its extreme on several truncated CenB haplotypes from which both KIR2DP1 and KIR2DL1 have been deleted ( Hilton et al. 2017a ; Pyo et al. 2013 ). Across human populations, the frequency of KIR B haplotypes correlates with the frequency of HLA-C2 ( Hiby et al. 2004 ). By extension, populations with high C2 frequency are replete with weak or inactive HLA-C2 receptor variants. This phenomenon is well illustrated by the KhoeSan of Southern Africa, a population having one of the highest known frequencies of C2 ( Gonzalez-Galarza et al. 2011 ) who have evolved two unique KIR2DL1 allotypes, neither of which has a normal functional interaction with C2. The first of these switched its specificity from C2 to C1 and the second, although able to bind to C2, has no connection to the inhibitory signaling apparatus ( Hilton et al. 2015b ). In analogous fashion, attenuated or inactivated C1-receptors are found in populations with high frequencies of HLA-C1. As an example, the Yucpa Amerindians, who have an unusually high frequency of C1 + HLA-C*07:02 evolved two KIR2DL3 variants, one with attenuated binding of C1 and the other that cannot bind C1 ( Gendzekhadze et al. 2009 ). Thus, inhibitory lineage III KIR appear to act as a buffer that respond to changes in HLA-C frequency. It is likely that this mechanism acts to maintain appropriate NK cell function, despite changes in HLA-C epitope frequency. That the binding affinity, cell-surface expression and signaling capacity of human inhibitory HLA-C receptors appear malleable ( Table 1 ), indicates they are well-suited to this purpose. NK cells and CD8 T cells play complementary roles in the cellular immune response. Whereas CD8 T lymphocytes kill cells presenting particular non-self peptides bound by HLA class I, NK cells kill cells with a deficiency of surface HLA class I. Reflecting these different roles, the T-cell receptor (TCR), the antigen receptor of CD8 T-cells, is exquisitely sensitive to changes in the bound peptide repertoire ( Garcia and Adams 2005 ). By contrast, KIR, the major variable receptors of NK cells, are less finely tuned, being able to recognize a broader range of peptide-HLA combinations ( Boyington and Sun 2002 ). Despite the broader specificity, pioneering studies in this field suggested that patterns of NK-cell mediated cytotoxicity were influenced by the nature of the peptide bound to HLA class I ( Malnati et al. 1995 ; Peruzzi et al. 1996 ; Storkus et al. 1992 ). Although KIR are not peptide specific like the TCR, there is considerable evidence that they exhibit a sophisticated peptide selectivity. There is now an increasing body of work exploring its extent and functional consequences ( Table 2 ). KIR peptide selectivity was first suggested by studies in which mutations within the peptide binding cleft of HLA class I were sufficient to change the cytotoxic response of NK cell clones to target cells ( Malnati et al. 1995 ; Peruzzi et al. 1996 ). Subsequent studies using amino-acid substituted peptide variants showed how specific peptides could modulate KIR2DL1 and KIR2DL2 recognition of HLA-C ( Boyington et al. 2000 ; Rajagopalan and Long 1997 ; Zappacosta et al. 1997 ). These studies highlighted the formative roles for the amino acid residues at P7 and P8 of the bound peptide. Large, negatively charged amino acids reduced or abrogated KIR binding and exposed target cells to NK-cell mediated lysis. The structural basis for this selectivity became clear from the crystallographic structure of the KIR2DL2/HLA-C*03:04 complex, which showed how KIR2DL2 binds over the C-terminal part of the bound peptide, making direct contacts with residues P7 and P8 ( Boyington et al. 2000 ). The structures of KIR2DS2 bound to HLA-A*11:01 and KIR2DL1 bound to HLA-C*04:01 show broadly similar binding modes, though they exhibit subtle differences in peptide accommodation ( Fan et al. 2001 ; Liu et al. 2014 ). Unlike KIR2DL2 and 2DS2, KIR2DL1 does not make any direct contact with bound peptide, despite its demonstrated sensitivity to mutation at P8 ( Fan et al. 2001 ; Rajagopalan and Long 1997 ). Recent studies point to differences in peptide selectivity within the various inhibitory KIR that recognize HLA-C ( Cassidy et al. 2015 ; Sim et al. 2017 ). Consistent with the structural picture, KIR2DL2/3 shows greater peptide selectivity than KIR2DL1 ( Sim et al. 2017 ). Supporting this finding, our examination of endogenously processed peptides from C1 + HLA-C and HLA-B showed that <25% of the peptides are strong KIR2DL3 ligands ( Hilton et al. 2017b ). The peptide selectivity of KIR2DL2/3 was further highlighted by the observation that peptide positions other than P7 and P8 have potential to influence the recognition of C1 + HLA-C ( Hilton et al. 2017b ; Sim et al. 2017 ). That KIR2DL3 appears as the receptor with the most peptide selectivity provides a compelling rationale for the observation that KIR2DL3 binding to C1 + HLA-C is weaker than KIR2DL1 binding to C2 + HLA-C ( Moesta et al. 2008 ). These studies show that KIR display peptide selectivity, but retain the capacity to bind a range of different peptide-HLA combinations. Studies examining the KIR reactivity of single peptide-HLA complexes were crucial in establishing the peptide selectivity of KIR binding. However, they were not designed to inform how a mixture of competing peptides, as occurs in vivo , acts to modulate NK cell reactivity. Experiments to compare the effects of peptides that bind weakly or strongly to the same HLA-C allotype are now addressing this question. A key discovery from such studies was that some weak KIR ligands compete effectively with strong KIR ligands. This phenomenon is termed peptide antagonism ( Cassidy et al. 2015 ; Fadda et al. 2010 ). Such competition reduces KIR-mediated inhibition ( Borhis et al. 2013 ; Fadda et al. 2010 ) to favor target cell lysis. This provides an alternative mechanism by which the innate immune system can respond to changes in the peptide repertoire bound by HLA-C. Peptide antagonism is hypothesized to be important during viral infection when subtle changes in the peptide repertoire occur in the absence of HLA class I down-regulation ( Fadda et al. 2010 ). During a viral infection, interactions between KIR and HLA-C allow NK cells to mount an effective immune response in complementary ways. Down-regulation of HLA-C triggers cytotoxicity by a missing-self response, whereas presentation of viral peptide antigens that are weak KIR ligands triggers cytotoxicity through peptide antagonism ( Figure 2 ). To this extent, KIR peptide selectivity can be advantageous to the host in responding to viral infection. Conversely, viral peptides that form strong ligands for inhibitory KIR may undermine the immune response by preventing NK-cell mediated killing of infected cells ( Figure 2 ). Such “escape mutants” were first proposed by Alter et al. , who found that new mutations in the HIV-1 sequence enhanced the binding of KIR2DL2 to HIV-1 infected CD4 + T-cells ( Alter et al. 2011 ). This enhanced binding correlated with reduced antiviral activity of NK cells and suggested that direct NK-cell mediated immune pressure on HIV-1 had selected for the new mutations. A peptide derived from the variant HIV-1 p24 Gag epitope reduces the cytotoxicity mediated by KIR2DL3 + NK cells, providing direct evidence for selection of a Gag peptide that prevents NK-cell response to HIV-infected cells ( Holzemer et al. 2015 ; van Teijlingen et al. 2014 ). This Gag-derived variant epitope is well expressed by HLA-C*03:04 and enhances binding of KIR2DL3, probably because of the T for V substitution at P8 of the variant peptide ( Holzemer et al. 2015 ). Taken together, it is clear that KIR bind numerous peptide-HLA combinations, but also retain a degree of peptide selectivity that has functional consequences for NK-cell mediated innate immunity. Further, the degree of KIR peptide selectivity appears finely balanced: high selectivity promotes strong immunity through peptide antagonism but also allows a rapidly-evolving virus to escape detection by production of strong inhibitory peptide. Low KIR peptide selectively reduces the potential for peptide antagonism but also makes viral escape less likely. Despite the recent advances, further investigation will be needed to understand how KIR respond to the complex peptide repertoires that result from HLA class I polymorphism, viral infection and malignant transformation. Two lines of evidence have suggested that the activating lineage III KIR recognize epitopes of HLA class I. One is their extensive sequence homology with the inhibitory lineage III KIR that recognize HLA-C ( Abi-Rached and Parham 2005 ); the other is the large number of epidemiologic studies that have correlated activating lineage III KIR and their putative ligands with a spectrum of infectious, auto-immune and reproductive diseases (summarized in Table 3 ). Despite these implications, ligands for the five activating lineage III KIR have, in general, been harder to identify than those for their inhibitory counterparts. Contributing to this relatively slow progress is their overall lower affinity for HLA class I ( Graef et al. 2009 ; Hilton et al. 2012 ; Hilton et al. 2015a ; Moesta et al. 2010 ; Older Aguilar et al. 2011 ; Stewart et al. 2005 ), a narrower specificity for HLA class I and an increased peptide selectivity ( Graef et al. 2009 ; Liu et al. 2014 ; Stewart et al. 2005 ). A further obstacle has been a dearth of cellular assays that can be used to probe the binding specificity of activating KIR on NK cells, and the functional response they produce. A reporter system, in which signaling through a modified activating KIR produces intracellular expression of a fluorescent protein, is an approach with potential to solve this problem ( van der Ploeg et al. 2017 ). KIR2DS1 recognition of HLA-C2 targets has been shown in a number of experimental systems and it is the activating KIR for which the ligand specificity is most clearly defined ( Table 4 ) ( Hilton et al. 2015a ; Moesta et al. 2010 ; Pende et al. 2009 ; Stewart et al. 2005 ; van der Ploeg et al. 2017 ). KIR2DS1 has an affinity for C2 that is significantly lower (~50%) than that of inhibitory KIR2DL1 ( Hilton et al. 2012 ; Hilton et al. 2015a ; Moesta et al. 2010 ). This reduced avidity stems largely from the threonine to lysine mutation at position 70, in the D1 domain ( Hilton et al. 2012 ). In contrast, the presence at position 45 of tyrosine in KIR2DS2 and phenylalanine in KIR2DL2, reduces the avidity of KIR2DS2 for C1 + HLA-C to a barely detectable level ( Moesta et al. 2010 ; Saulquin et al. 2003 ; Winter et al. 1998 ). This low avidity may be compounded by the considerable peptide selectivity of KIR2DS2, which is apparent from the structural analysis of KIR2DS2 bound to HLA-A*11:01 ( Liu et al. 2014 ). KIR2DS4 also recognizes HLA-A*11:02 in a peptide dependent way ( Graef et al. 2009 ), the result of a proline 71, valine 72 motif, which was introduced into KIR2DS4 by gene conversion with KIR3DL2 prior to the divergence of humans from chimpanzees ( Graef et al. 2009 ). The reported peptide selectivity of activating KIR may enable NK cells to eliminate infected cells through recognition of viral peptides bound by HLA class I. Supporting this hypothesis, KIR2DS1 has been shown to recognize a viral-induced ligand on CMV infected HLA-C2 + human fibroblasts ( van der Ploeg et al. 2017 ). However, the mechanistic basis for this recognition is not fully understood and could result from changes in HLA-C glycosylation, co-presentation of HLA-C with a viral ligand or recognition of HLA-C open conformers that lack associated peptide or β 2 -microglobulin ( van der Ploeg et al. 2017 ). Several activing KIR, including KIR2DS4, have been shown to recognize an open conformer of the non-classical HLA-F ( Goodridge et al. 2013 ). A similar mechanism may underlie the recent observation that KIR2DS2 recognizes cancer cell lines in a β 2 -microglobulin independent fashion ( Thiruchelvam-Kyle et al. 2017 ). The direct recognition of virally-induced ligands by activating NK cell receptors has been observed in mice ( Arase et al. 2002 ; Smith et al. 2002 ), but is yet to be confirmed in the human system. Together, these studies suggest that activating KIR recognize altered forms of HLA class I that may be up-regulated during viral infection or malignant transformation. Activating KIR are expressed and appear functional on uterine NK cells ( Hiby et al. 2010 ; Kennedy et al. 2016 ; Xiong et al. 2013 ). Further, epidemiologic studies are consistent with the activating KIR playing a role in placental development ( Hiby et al. 2004 ; Hiby et al. 2010 ; Hiby et al. 2014 ; Nakimuli et al. 2015 ). The genes of the telomeric KIR B haplotype ( TelB ) have consistently been associated with protection against the development of pre-eclampsia and other disorders of pregnancy that result from insufficient invasion by extravillous trophoblast ( Hiby et al. 2004 ; Hiby et al. 2010 ; Nakimuli et al. 2015 ). This protective effect is more marked in pregnancies where the fetus expresses HLA-C2, suggesting that a C2-specific KIR encoded by TelB mediates the protection. In women of European ancestry this KIR is KIR2DS1 ( Hiby et al. 2010 ), whereas in women of African ancestry it is KIR2DS5 ( Nakimuli et al. 2015 ). Previously not known to bind any HLA class I ligand, the protective *006 allotype of KIR2DS5 has recently been shown to bind weakly to C2 + HLA-C (Blokhuis et al 2017, submitted). By contrast, KIR2DS5*002, which showed no correlation with protection, does not bind to bind any HLA class I. This result highlights the extent to which allelic the polymorphism of an activating KIR can contribute different ligand binding properties. Although considerably less polymorphic than the inhibitory lineage III KIR, similar changes in binding affinity have been observed for KIR2DS1 allotypes ( Hilton et al. 2015a ). The extent to which peptide selectivity of activating KIR is important in the context of reproduction has yet to be determined. However, given that HLA-C expressed on extravillous trophoblast may sample a unique proteome, further studies of its effect are warranted.