KIR2DS5 allotypes that recognize the C2 epitope of HLA‐C are common among Africans and absent from Europeans

Abstract Introduction KIR2DS5 is an activating human NK cell receptor of lineage III KIR. These include both inhibitory KIR2DL1, 2 and 3 and activating KIR2DS1 that recognize either the C1 or C2 epitope of HLA‐C. In Europeans KIR2DS5 is essentially monomorphic, with KIR2DS5*002 being predominant. Pioneering investigations showed that KIR2DS5*002 has activating potential, but cannot recognize HLA‐A, ‐B, or ‐C. Subsequent studies have shown that KIR2DS5 is highly polymorphic in Africans, and that KIR2DS5*006 protects pregnant Ugandan women from preeclampsia. Because inhibitory C2‐specific KIR2DL1 correlates with preeclampsia, whereas activating C2‐specific KIR2DS1 protects, this association pointed to KIR2DS5*006 being an activating C2‐specific receptor. To test this hypothesis we made KIR‐Fc fusion proteins from all ten KIR2DS5 allotypes and tested their binding to a representative set of HLA‐A, ‐B and ‐C allotypes. Results Six African‐specific KIR2DS5 bound to C2+HLA‐C but not to other HLA class I. Their avidity for C2 is ∼20% that of C2‐specific KIR2DL1 and ∼40% that of C2‐specific KIR2DS1. Among the African C2 receptors is KIR2DS5*006, which protected a cohort of pregnant Ugandans from pre‐eclampsia. Three African KIR2DS5 allotypes and KIR2DS5*002, bound no HLA‐A, ‐B or ‐C. As a group the C2‐binding KIR2DS5 allotypes protect against pre‐eclampsia compared to the non‐binding KIR2DS5 allotypes. Natural substitutions that contribute to loss or reduction of C2 receptor function are at positions 127, 158, and 176 in the D2 domain. Conclusions KIR2DS5*005 has the KIR2DS5 consensus sequence, is the only allele found at both centromeric and telomeric locations of KIR2DS5, and is likely the common ancestor of all KIR2DS5 alleles. That KIR2DS5*005 has C2 receptor activity, points to KIR2DS5*002, and other allotypes lacking C2 receptor function, being products of attenuation, a characteristic feature of most KIR B haplotype genes. Alleles encoding attenuated and active KIR2DS5 are present in both centromeric and telomeric locations.


Introduction
Killer cell immunoglobulin-like receptors (KIR) are expressed by subsets of human NK cells, a subpopulation of lymphocytes that contributes to innate immunity, adaptive immunity and reproduction [1,2]. The target cell ligands recognized by KIR are epitopes of HLA-A, -B and -C. These epitopes are defined by alternative sequence motifs at residues 76-83 of the a 1 domain. The human KIR family comprises four phylogenetic lineages, of which lineage III includes all KIR that recognize the C1 and C2 epitopes of HLA-C. These epitopes are defined by dimorphism at position 80 of HLA-C, where asparagine confers C1 specificity and lysine confers C2 specificity [3].
KIR2DL1 is an inhibitory receptor that is highly specific for C2 whereas KIR2DL2 and KIR2DL3 are inhibitory receptors that principally recognize C1 [4]. Moreover, KIR2DL2, and to lesser extent KIR2DL3, cross-react with some C2 þ HLA-C [5,6]. In addition to these much-studied inhibitory lineage III receptors, there are five activating lineage III receptors: KIR2DS1, 2DS2, 2DS3, 2DS4, and 2DS5. In comparison to the inhibitory receptors, these activating receptors are not as well characterized [7,8]. Only KIR2DS1 has been shown to recognize HLA-C with specificity like that of an inhibitory KIR. Like inhibitory KIR2DL1, KIR2DS1 has methionine 44 and is specific for the C2 epitope. But the avidity of KIR2DS1 for C2 is about half that of KIR2DL1 [7,8].
Epidemiological studies have correlated the presence or absence of the KIR2DS1 gene with several pregnancy syndromes [9][10][11]. In these and other disease association studies, it has proved useful to divide the numerous KIR haplotypes into two groups, KIR A and KIR B, based upon their content of activating KIR genes. KIR A haplotypes can have either no activating KIR genes, a functional form of the KIR2DS4 gene or a non-functional form of the KIR2DS4 gene. In contrast, KIR B haplotypes can have many different combinations of the five KIR2DS, as well as KIR3DS1. Pregnant women who have a KIR A/A genotype are at greater risk for pre-eclampsia than pregnant women having an A/B or B/B genotype. This correlation indicates that B haplotypes protect against pre-eclampsia. For pregnancy of A/A mothers, the risk increases further if the fetus has C2 + HLA-C, especially if the mother lacks C2 + HLA-C. This correlation strongly implies that interaction of fetal C2 + HLA-C with KIR2DL1 is the cause of the increased risk [9]. Study of European cohorts has shown that the protective effect of a maternal B haplotype is mediated by KIR2DS1, the activating C2 receptor [10]. Thus the interaction between fetal C2 and activating KIR2DS1 can counter that of fetal C2 with inhibitory KIR2DL1, and in this way reduce the likelihood of pre-eclampsia [12].
Similar study of a Ugandan cohort of pregnant women correlated a different activating receptor, KIR2DS5, with protection from pre-eclampsia. Of 11 KIR2DS5 alleles present in this cohort of sub-Saharan Africans, KIR2DS5 Ã 006 was associated with significant protection [11]. That KIR2DS5 Ã 006 appears to counter the interaction between C2 and KIR2DL1, implies that KIR2DS5 Ã 006 is an activating receptor that recognizes the C2 epitope of HLA-C.
Whereas KIR2DS5 is highly polymorphic in Africans [13] and African-Americans [14] that is not the case for Europeans and other populations outside Africa. In these populations the KIR2DS5 Ã 002 allotype dominates [15]. Also present in the cohort of Ugandan women studied by Nakimuli et al. [11], the KIR2DS5 Ã 002 allotype provided no protection against pre-eclampsia, consistent with previous cellular and molecular analyses that failed to detect any functional or molecular interaction of KIR2DS5 Ã 002 with C1 + HLA-C, C2 + HLA-C, or any HLA-A or HLA-B variant [8,16]. With this background, the aim of our study was to test the hypothesis that KIR2DS5 Ã 006 differs from KIR2DS5 Ã 002, and is an activating receptor for the C2 epitope of HLA-C.
KIR2DS5 polymorphism modulates receptor avidity for C2 + HLA-C Ten of the eleven KIR2DS5 allotypes present in the Ugandan cohort studied by Nakimuli et al. [11] are distinguished by substitutions in the extracellular part of the receptor. Of the eight dimorphisms that distinguish the 10 allotypes, six are in the D2 domain, one is in the D1 domain and one is in the stem (Fig. 2). KIR-Fc corresponding to the ten KIR2DS5 allotypes were made and compared for binding to HLA-A, -B, and -C. Positive reactions were observed only for KIR2DS5-Fc binding to C2 + HLA-C. The 10 KIR2DS5 allotypes divide into two groups: one group of six allotypes (2DS5 Ã 003, Ã 004, Ã 005, Ã 006, Ã 007, and Ã 008), that bound C2 + HLA-C to a level that is 15-20% of the KIR2DL1 Ã 003 binding and a second group of four allotypes (2DS5 Ã 002, Ã 009, Ã 010, and Ã 011) that bound C2 + HLA-C to an extent that was less than 8% of the KIR2DL1 Ã 003 binding (Fig. 2). In the Ugandan cohort we see that 66.5% of the KIR2DS5 allotypes recognize C2 + HLA-C and 33.5% of them do not (Fig. 3). Thus in Africans a majority of the KIR2DS5 allotypes are active C2 + HLA-C receptors, whereas in Europeans almost none of them are.
We repeated the analysis described by Nakimuli et al. [11], but with the KIR2DS5 allotypes clustered according to their binding avidities ( Table 1). The C2-binding KIR2DS5 allotypes are seen to protect against pre-eclampsia compared to the non-binding allotypes (p ¼ 0.01801, OR 0.68). Thus, our functional analysis helps resolve which particular KIR gene in the African-specific cB01 region protects Ugandan women against pre-eclampsia. Each data point represents the binding to a different HLA class I allotype. For each HLA-coated bead, the binding data obtained with KIR-Fc were normalized to that of the W6/32 antibody, which reacts equivalently with all HLA-A, -B and -C allotypes (calculation detailed in Materials and Methods section). Four independent binding assays were performed for each KIR-Fc and a representative experiment is shown. The mean binding value is indicated by the horizontal bar. The binding of KIR2DS5 Ã 006 to C2 þ HLA-C is significantly higher than that of KIR2DS5 Ã 002 (two-tailed paired Student's t-test, p ¼ 0.0076). The statistical analysis used mean binding values from four independent assays (see Figure S1). (B) Binding of the KIR2DL3 Ã 001, 2DS5 Ã 002 and 2DS5 Ã 006-Fc fusion proteins to each of the seven C2 + HLA-C allotypes is shown normalized to that of KIR2DL1 Ã 003. Values shown are representative of four independent binding experiments for each KIR-Fc.
As a group the six KIR2DS5 Ã allotypes that bind C2 are associated with protection from pre-eclampsia, but in testing the allotypes individually only KIR2DS5 Ã 006 is significantly protective [11]. A key question is why the other C2-binding allotypes (2DS5 Ã 003, Ã 004, Ã 005, Ã 007, and Ã 008) are not associated with protection. One possibility is that there are differences in the expression of KIR2DS5 allotypes, which change the amount of receptor at the NK cell surface or the frequency of uterine NK cells expressing the receptor. A second possibility is that other KIR2DS5 allotypes do provide protection but because of their low frequencies their effects did not reach statistical significance in the study of Nakimuli et al. [11]. Larger cohorts will address this question as well as the contributions to pre-eclampsia of different C2 + HLA-C ligands in the fetus.
KIR2DS5 Ã 005 has the consensus sequence of all KIR2DS5 alleles. It is also the only allele found at both the centromeric and telomeric locations of the KIR2DS5 gene within the KIR locus. These properties point strongly to KIR2DS5 Ã 005 being the common ancestor of all other KIR2DS5 alleles. That KIR2DS5 Ã 005 is a high-binding allotype implies that KIR2DS5 Ã 002 and the other low-binding receptors are derived forms, having acquired substitutions that reduced their avidity for C2 + HLA-C. Phenylalanine 127 and arginine 158 correlate with conversion from high to low binding. Comparison of KIR2DS5 Ã 002 and KIR2DS5 Ã 009 shows how threonine 176 augments the reduction caused by arginine 158 (Fig. 2). In addition to this affect on C2 binding, Steiner et al. [23] showed that arginine 158, the only difference between KIR2DS5 Ã 002 and KIR2DS5 Ã 005, causes intracellular retention of KIR2DS5 Ã 002, as well as KIR2DS5 Ã 009. They attribute this effect to arginine 158 interfering with glycosylation of asparagine 157 [23]. Despite these deleterious properties, cells expressing KIR2DS5 Ã 002 can transmit activating signals when incubated with an antibody that binds KIR2DS5 [16].
Comparison of the organization of the KIR locus in humans and other hominoids shows that lineage III KIR genes originated in the centromeric region. During human evolution, about 1.7 million years ago, a duplication of the centromeric progenitor of the KIR2DS5 and KIR2DS3 genes introduced a copy of the KIR2DS3/5 progenitor into the telomeric region [24]. Subsequent divergence resulted in the distinctive KIR2DS3 and KIR2DS5 allelic lineages that have an average of 18 amino acid differences compared to an average of two differences between pairs of sequences in the same lineage. The substitutions that distinguish KIR2DS3 and KIR2DS5 are spread throughout the sequence, suggesting that the two KIR2DS3/5 genes diverged independently. Having the consensus sequence points to KIR2DS5 Ã 005 as the ancestor of all extant KIR2DS5 alleles. The location of KIR2DS5 Ã 005 in both the centromeric and telomeric intervals could be interpreted as evidence for KIR2DS5 Ã 005 being the ancestor of all KIR2DS3 and KIR2DS5. However, the large number of differences distinguishing these allelic lineages indicates that the dual position of KIR2DS5 Ã 005 is more likely the product of a subsequent unequal recombination event that occurred after the divergence of the KIR2DS3 and KIR2DS5 allelic lineages. Of the five KIR2DS5 allotypes encoded in the centromeric region, four recognize C2 + HLA-C, whereas only three of the six KIR2DS5 allotypes Figure 2. KIR2DS5 allotypes differ in their capacity to recognize C2 þ HLA-C. Shown are the relative frequencies of the ten KIR2DS5 allotypes observed in the cohort of pre-eclampsia patients studied by Nakimuli et al. [11], [Frequencies are calculated from KIR2DS5 + individuals only and are not the allele frequencies in the patient population, many of whom lack KIR2DS5]. Listed for each allotype are the amino-acid substitutions that distinguish their extracellular domains (D1 and D2 and Stem) and the binding of their corresponding KIR-Fc to seven C2 þ HLA-C allotypes. KIR2DS5 Ã 005, the most frequent allotype in this cohort, is set as the consensus sequence with blank boxes indicating sequence identity to the consensus. Shown on the right is the mean binding of each KIR2DS5-Fc to each C2 þ HLA-C, normalized to that of 2DL1 Ã 003. The dashed vertical line shows KIR2DL3 Ã 001-Fc binding to the same allotypes. Figure 3. The majority of African KIR2DS5 allotypes recognize C2 þ HLA-C. Shown are the relative frequencies of the KIR2DS5 allotypes that do, or do not, bind C2 þ HLA-C. The frequencies are from the cohort studied by Nakimuli et al. [11] and represent relative frequencies of KIR2DS5 allotypes in KIR2DS5 þ individuals in this population. The KIR2DS5 alleles distribute between locations in centromeric and telomeric regions of the KIR locus. Shown is the distribution between the centromeric and telomeric regions of the KIR locus of the alleles encoding KIR2DS5 allotypes that, do or do not, bind C2 þ HLA-C. KIR2DS5 Ã 005 has the highest frequency and is the only allele found in both genomic regions. encoded in the telomeric region recognize C2 + HLA-C (Fig. 3). Thus there has been increased attenuation of telomeric KIR2DS5 alleles than centromeric KIR2DS5 alleles. Of the eight substitutions that distinguish other KIR2DS5 allotypes from KIR2DS5 Ã 005 (the ancestral allele), three (F127, R158, and T176) caused major loss of ligand binding, two had little effect (R1 and H182) and three gave modest increases in ligand binding (N123, T154, and K216). In their comparison of eight KIR2DS5 allotypes, Steiner et al. [25] showed that KIR2DS5 Ã 002 and KIR2DS5 Ã 009 have low cellsurface expression, because they are largely retained inside cells, whereas KIR2D5 Ã 003, Ã 004, Ã 005, Ã 006, Ã 007, and Ã 008 have higher cell-surface expression. This difference correlates with the presence of arginine 158 in KIR2DS5 Ã 002 and KIR2DS5 Ã 009 (Fig. 2). Acquisition of this residue has had two critical effects, reducing cell-surface expression and reducing avidity for C2. It has been proposed that arginine 158 exerts these effects by interfering with the glycosylation at asparagine 157. Asparagine 123, which distinguishes KIR2DS5 Ã 003 from KIR2DS5 Ã 005, creates a glycosylation site and increases the level of cell-surface expression of KIR2DS5 Ã 003. Thus the increased expression could be due to the substitution of serine for asparagine at position 123, to the carbohydrate attached to asparagine 123 or combination of these two factors. Cellsurface expression of KIR2DS5 Ã 002 is also increased by having asparagine at position 123 instead of serine [25].
In non-African populations, KIR2DS5 Ã 002 is part of a conserved, telomeric KIR haplotype that consists of KIR2DL4 Ã 00501, 3DS1 Ã 013, 2DL5A Ã 001, 2DS5 Ã 002, 2DS1 Ã 002, and 3DL2 Ã 00701. This conservation is well illustrated by the IHWG panel of HLA homozygous cells lines that are predominantly of European origin. Although homozygous for HLA, these cells are almost all heterozygous for the KIR locus. Their KIR haplotypes have been determined at allele-level resolution by complete nucleotide sequencing [26]. Among the 194 KIR haplotypes of the 97 cell lines, are 31 that contain KIR2DS5 Ã 002. Of these, 28 have the conserved telomeric haplotype. The other three haplotypes differ only from the conserved haplotype by their KIR3DL2 allele. One additional telomeric haplotype contains a KIR2DS5 gene. This haplotype is the same as the conserved haplotype with the exception of the KIR2DS5 allele which is KIR2DS5 Ã 015. This allele differs by a single substitution that results in substitution of aspartate by asparagine at position 271. Thus all haplotypes with KIR2DS5 Ã 002 have KIR2DS1 Ã 002 as the neighboring gene, which encodes an activating C2 receptor that is stronger than any of the KIR2DS5 allotypes. This raises the possibility that emergence of KIR2DS1 directly led to the attenuation of KIR2DS5, as is most clearly manifest for KIR2D5 Ã 002.
In a variant of the conserved telomeric haplotype, KIR2D5 Ã 002 is replaced by KIR2DS3 Ã 002 and KIR2DL5A Ã 001 is replaced by KIR2DL5A Ã 005. KIR2DS3 has sequence similarity with KIR2DS5 and also encodes KIR with threonine 44. There is considerable evidence for the attenuation of KIR2DS3 [27] which is poorly expressed at the cell surface. This too could have been caused by emergence of KIR2DS1. The third human KIR that encodes threonine 44 is the inactivated KIR2DP1 gene, which once encoded two lineages of inhibitory KIR allotypes, one having lysine 44 and specificity for the C1 epitope, the other having threonine 44 and specificity for the C2 epitope [28]. The complete or partial demise of the human KIR with threonine 44, could have been driven by the emergence of the stronger KIR2DL1 and KIR2DS1 C2-specific receptors with methionine 44. KIR with threonine 44 have been found only in the human species and may represent an evolutionary intermediate that allowed the human KIR system to recover from the loss of KIR diversity that accompanied human speciation [28].
The rarity of the conserved telomeric KIR2DS5 Ã 002 haplotype in anthropologically well characterized African populations (Nemat-Gorgani et al., unpublished data and [13]) and its prevalence in European populations, raises the possibility that modern Europeans acquired this haplotype from archaic Europeans, such as Neandertals, a phenomenon that we previously considered in the context of the KIR3DS1 Ã 013 component of the haplotype [29]. A major benefit of the haplotype was likely conferred by KIR2DS1 in reducing the incidence of pre-eclampsia and related pregnancy syndromes [9]. On the other hand, study of European pregnancies has shown that excessive activation caused by paternal C2 þ HLA-C interacting with maternal KIR2DS1 can lead to babies with high birth-weight and the potential complication of obstructed labor, which in the absence of surgical intervention can cause death of both mother and child [30]. Because KIR2DS5 Ã 006 has a lower avidity for C2 than KIR2DS1, this activating C2 receptor of Africans is predicted to have a reduced propensity for excessive activation than KIR2DS1 in Europeans. Indeed, as has recently been reviewed [12], Africans have fewer high birth-weight babies than Europeans with the result that preeclampsia and other conditions associated with low birth weight, are more prevalent than obstructed labor.

Ethics statement
The experiments reported here involved no animal or human subjects. Results presented in Table 1 regarding the Ugandan cohort are a reanalysis of previously published data [11].

Sequences of KIR2DS5 alleles and proteins
The amino acid sequences of 12 KIR2DS5 allotypes are deposited in the Immunoreceptor Polymorphism Database (IPD) [31]. These sequences were first aligned automatically in Seaview 4.4.0 [32] and at positions where the alignment could not be resolved it was manually aligned within the same program. KIR-Fc fusion proteins, corresponding to 10 of the KIR2DS5 allotypes, were made. One KIR-Fc represented both KIR2DS5 Ã 006 and KIR2DS5 Ã 012, because they have identical D1 and D2 domains, differing only by a single amino acid substitution in the signal sequence. The KIR2DS5 Ã 001 sequence was not used to make a KIR-Fc because it is unlikely to represent a natural KIR2DS5 allotype, does not fold properly and is retained inside cells [23,25]. Although it was the first KIR2DS5 sequence determined [33] subsequent analyses failed to identify any individual or cell line that types for the KIR2DS5 Ã 001 allele [34,35]. That no confirmation of KIR2DS5 Ã 001 has occurred in the 20 years since it was first reported [33] [17]. The integrity of KIR-Fc fusion protein folding was assessed by flow cytometry as described [17]. The Sf9 and Hi5 cells were kindly provided by Chris Garcia, Stanford University. KIR-Fc fusion proteins corresponding to KIR2DS5 allotypes and site-directed mutants, as well as the KIR2DL1 Ã 003 and KIR2DL3 Ã 001 controls, were tested for binding to a panel of 97 microbeads, each coated with one of 31 HLA-A, 50 HLA-B, and 16 HLA-C allotypes (LabScreen Single-Antigen beads lot #8, One Lambda, Kittridge, CA). To account for differences in the amount of HLA class I protein coating each of the 97 beads, the binding of KIR-Fc fusion proteins was normalized to that of W6/32, a monoclonal antibody detecting an epitope shared by all HLA class I. Normalized values were calculated using the formula: (specific binding-bead background)/(W6/32 binding-bead background). As a comparison for KIR2DS5 binding, we used our published binding data for KIR2DS1 Ã 002, the known activating receptor that is specific for C2 [4,8].

Genetic analysis
Samples of DNA from a case-control study of pre-eclampsia involved 738 pregnant women at Mulago Hospital, Kampala in Uganda were typed for KIR and HLA-C variants including for presence/absence of KIR2DS5 alleles [11,37]. Categorical data was analyzed with chi-square and Fisher's exact test with two-tailed mid-p adjustment. A p-value of 0.05 was considered to be statistically significant. The magnitude of the effect was estimated by conditional maximum likelihood estimate of Odds Ratio (OR) for the mid-p exact test and their 95% confidence intervals (CI).

SUPPORTING INFORMATION
Additional supporting information may be found in the online version of this article at the publisher's web-site Figure S1. Binding of KIR2DS5-Fc to HLA-C allotypes is specific to C2 epitopes. (A-B) Individual KIR2DS5-Fc were incubated with a panel of 97 microbeads, each coated with one of 31 HLA-A, 50 HLA-B and 16 HLA-C allotypes. KIR2DS5-Fc binding normalized to W6/32 binding (described in Materials and Methods) is shown in (A) C2 þ HLA-C and (B) C1 þ HLA-C and C1 þ HLA-B. Data is displayed as mean AE SD and comes from a minimum of two independent binding assays for each KIR2DS5-Fc. No binding was observed to any HLA-A or the remaining 48 C1 -HLA-B allotypes (data not shown).