Full correspondence: Dr. Christelle Retiere, Etablissement Français du Sang Pays de la Loire, EA4271 “Immunovirologie et Polymorphisme Génétique”, 34 boulevard Jean Monnet, 44011 Nantes cedex 01, France
Additional correspondence: Dr. Katia Gagne, Etablissement Français du Sang Pays de la Loire, EA4271 “Immunovirologie et Polymorphisme Génétique”, 34 boulevard Jean Monnet, 44011 Nantes cedex 01, France. e-mail: email@example.com
NK-cell functions are regulated by many activating and inhibitory receptors including KIR3DL1. Extensive allelic polymorphism and variability in expression can directly alter NK-cell phenotype and functions. Here we investigated the KIR3DL1+ NK-cell repertoire, taking into account the allelic KIR3DL1/S1 polymorphism, KIR3DL1 phenotype, and function. All 109 studied individuals possessed at least one KIR3DL1 allele, with weak KIR3DL1*054, or null alleles being frequently present. In KIR3DL1high/null individuals, we observed a bimodal distribution of KIR3DL1+ NK cells identified by a different KIR3DL1 expression level and cell frequency regardless of a similar amount of both KIR3DL1 transcripts, HLA background, or KIR2D expression. However, this bimodal distribution can be explained by a functional selection following a hierarchy of KIR3DL1 receptors. The higher expression of KIR3DL1 observed on cord blood NK cells suggests the expression of the functional KIR3DL1*004 receptors. Thus, the low amplification of KIR3DL1high, KIR3DL1*004 NK-cell subsets during development may be due to extensive signaling via these two receptors. Albeit in a nonexclusive manner, individual immunological experience may contribute to shaping the KIR3DL1 NK-cell repertoire. Together, this study provides new insight into the mechanisms regulating the KIR3DL1 NK-cell repertoire.
The effector functions of NK cells are regulated by inhibitory and activating receptors, e.g. killer cell immunoglobulin-like receptors (KIR), which are specific for allotypic determinants shared by different HLA-class I molecules . In particular, HLA-Cw allotypes with Asn80 (C1 ligands) or Lys80 (C2 ligands) are, respectively, recognized by KIR2DL2/2DL3 and KIR2DL1. HLA-A and HLA-B allotypes with a Bw4 motif are recognized by KIR3DL1 whereas HLA-A3/A11 are recognized by KIR3DL2 [2, 3]. Lack of inhibitory KIR engagement can trigger alloreactive KIR NK-cell cytotoxicity only within functionally competent NK cells . Although the ligands and functions of inhibitory KIR receptors are well documented, this is not the case for activating KIR receptors and their ligands, except for KIR2DS1 which recognizes C2 ligands only in C2-adult individuals . Moreover, the activating receptor KIR3DS1, which segregates as an allele of KIR3DL1 , shares more than 97% sequence homology in its extracellular domain with the KIR3DL1 receptor. However, a functional interaction with HLA-A or HLA-B allotypes sharing the Bw4 public epitope has not been demonstrated in vitro [7, 8] despite a significantly higher frequency of KIR3DS1+ NK cells observed in Bw4+ than in Bw4− individuals .
KIR genes are located on chromosome 19q13.4. To date, 14 functional KIR genes have been characterized . Within the human population, genomic diversity of the KIR region is achieved on several levels. KIR gene content varies between individuals who can exhibit 7–14 inhibitory and activating KIR genes . Population studies have demonstrated two major KIR haplotypes: A and B . The A haplotypes correspond to 7 KIR genes, including KIR2DS4 as the only activating KIR gene. In contrast, B haplotypes are more diverse and are characterized by the presence of more than one activating KIR gene and the absence of the KIR2DS4 gene [13, 14]. KIR gene polymorphism is the largest contributor to KIR region diversity, with multiple alleles defined . While polymorphism is limited in the KIR2DL1 and KIR2DL2/L3 genes, it is much broader for KIR3DL1 and can even alter NK-cell functions [16, 17]. Indeed, depending on the KIR3DL1 allele present in a given individual, the level of KIR3DL1 expression differs on the NK-cell surface, i.e. null for KIR3DL1*004, low for KIR3DL1*005 and *007, and high for KIR3DL1*001, *002, *01502, and *008 alleles [18-20]. Two KIR3DL1 alleles can be transcribed in one NK-cell clone, and the proportion of clones expressing both alleles is high in comparison to clones expressing only one KIR3DL1 allele. However, based on the phenotypic pattern on all NK cells, only the nature of each allele is marked .
In addition to allelic polymorphism, the variability in expression levels of KIR3DL1 molecules on the NK-cell surface also varies depending on the presence or absence of autologous Bw4 molecules, promoter polymorphism, and DNA methylation [21-23]. Functional interaction between KIR3DL1 receptors and their HLA-Bw4 ligands are modulated by KIR3DL1 allotypes [24, 25] and the peptide content of HLA class I molecules . Moreover, KIR3DL1+ NK-cell cytotoxicity differs depending on KIR3DL1 allotypes and/or HLA-A, HLA-B Bw4+ targets [3, 27, 28]. From a clinical standpoint, KIR3DL1 allelic polymorphism has been correlated with HIV progression and viral load . We have previously shown that KIR3DL1/3DS1 gene disparities of donor/recipient pairs in hematopoietic stem cell transplantation have a significant impact on hematopoietic stem cell transplantation outcome .
Until now, the impact of KIR3DL1/3DS1 allele combinations on NK-cell phenotype and function has only been studied in one Japanese cohort not representative of Caucasians . However, the mechanisms shaping the KIR3DL1 repertoire remain ill-defined. The link between KIR3DL1 allelic polymorphism, levels of KIR3DL1 expression at the NK-cell surface, and frequency of KIR3DL1+ NK-cell subsets has yet to be clarified. This is necessary in order to better understand the mechanisms that shape the KIR3DL1+ NK-cell repertoire. In this study, we hypothesize that the high KIR3DL1 allelic polymorphism, including frequent null alleles observed in Caucasians, might impact the formation of the KIR3DL1 NK-cell repertoire. We therefore analyzed the KIR3DL1 NK-cell repertoire taking into account the KIR3DL1 allelic polymorphism, KIR3DL1/KIR3DS1 allele combinations, and phenotypic patterns such as frequency, the mean level of all KIR3DL1+ NK-cell subsets, the Bw4 environment, and the functional potential of these KIR3DL1+ NK-cell subsets.
High proportion of KIR3DL1high and KIR3DL1null alleles in French individuals
KIR3DL1/3DS1 allele combinations were investigated in AA (n = 34), AB (n = 67), or BB (n = 8) KIR genotyped individuals. Overall, 33 different KIR3DL1/3DS1 allele combinations were observed (Fig. 1A, central panel). AA and AB KIR genotyped individuals exhibited a higher KIR3DL1 allelic variability compared with BB KIR genotyped individuals. The KIR3DL1*001, *004 allele combination was the most frequent in AA KIR genotyped individuals (Fig. 1A). KIR3DL1*001, *002, *004, *005, *007, *008, *015 known alleles were the most frequent (Fig. 1A, right panel). In AA KIR genotyped individuals, the frequencies of KIR3DL1 alleles ranged from 1.61 (KIR3DL1*008, *015, *005, *009, *019) to 32.26% (KIR3DL1*004; Fig. 1A, right panel). In AB KIR genotyped individuals, the frequencies of KIR3DL1 alleles ranged from 0.71 (KIR3DL1*017, *009) to 17.73% (KIR3DL1*001; Fig. 1A, right panel). In our cohort, the highly expressed KIR3DL1*001 and null KIR3DL1*004 alleles were the most frequent, being observed at one or two doses in 39 (35%) and 32 (29%) KIR genotyped individuals, respectively (Fig. 1A, right panel). The presence of autologous HLA-A and/or HLA-B Bw4+ molecules did not affect both the KIR3DL1 allele and KIR3DL1/3DS1 allele combination frequencies (data not shown).
The frequency of KIR3DL1+ NK-cell populations and the KIR3DL1 expression level were further assessed using the KIR3DL1- and KIR3DS1-specific Z27 specific monoclonal antibody in 109 KIR3DL1 allele-typed individuals (Fig. 1B). To examine the potential effect of KIR3DL1/3DS1 allele combinations on NK KIR3DL1+ phenotype, individuals were divided into different groups based on null, low, and high KIR3DL1 expression, and weak expression of KIR3DS1. Seven different phenotypic patterns were identified: no binding pattern (KIR3DL1null,null/null genotypes), a unimodal pattern with weak intensity (KIR3DL1null/KIR3DS1), a unimodal pattern with low intensity (KIR3DL1low/low or KIR3DL1low/null), a unimodal pattern with high intensity (KIR3DL1high/high or KIR3DL1high/null), a bimodal pattern with weak and low intensities (KIR3DS1/KIR3DL1low), a bimodal pattern with weak and high intensities (KIR3DS1/KIR3DL1high), and a bimodal pattern with low and high intensities (KIR3DL1low/high). The bimodal expression patterns observed in KIR3DL1low/high and KIR3DS1/KIR3DL1high individuals indicates the presence of NK-cell subsets expressing each of the expected KIR3DL1 alleles, and are in agreement with the reported mutually exclusive expression of KIR3DL1 and KIR3DS1 . All combinations including KIR3DL1high, KIR3DL1low, KIR3DL1null, or KIR3DS1 were observed. The in-depth analysis of KIR3DL1 alleles and NK-cell phenotypes revealed the KIR3DL1 phenotypic profile for KIR3DL1*009 and KIR3DL1*01702 alleles which are, respectively, expressed at low and high levels on the NK-cell surface and highlight unusual KIR3DL1 phenotypic profiles in a low proportion of individuals (Fig. 1A). The distribution of all KIR3DL1/3DS1 allele combinations shows that KIR3DL1high alleles were abundant, and mainly associated with KIR3DL1null or KIR3DS1 (Fig. 1A and B).
The KIR3DL1*054 allele is frequently present and mainly associated with KIR3DS1
All KIR3DS1+ individuals for whom no KIR3DL1 allele was identified at the genotypic level had the KIR3DL1*054 allele , as illustrated in Figure 2A for two individuals and two well-characterized cell lines (13th International Workshop; WT51 and HOR) previously defined as KIR3DL1− . The KIR3DL1*001, *005 genotyped and characterized DEU cell line (13th International Workshop) was used as a positive control. Sequencing of all amplified products in nine 3DL1−, 3DS1+ genotyped individuals showed that the KIR3DL1*054 allele was unique, with specific T475, T550 and G560 nucleotides in exon 4 (data not shown). As described for KIR3DL1*004 [34, 35], no KIR3DL1 receptor was detected on the NK-cell surface for the only KIR3DL1*019, *004 individual studied (Fig. 2B). Alignment of amino acid substitutions of KIR3DL1*019 and KIR3DL1*054 compared with that of the KIR3DL1*004 allotype showed that the expected mature KIR3DL1*019 protein possesses the same amino acids involved in the intracellular retention of the null KIR3DL1*004 allotype, i.e. L86 and S182 (Fig. 2D). Only one amino acid in the D0 domain differed between KIR3DL1*019 and KIR3DL1*004. All KIR3DL1*054+ individuals identified in our cohort presented the KIR3DS1 allele (Fig. 1A). In KIR3DS1+ individuals, staining using the combination of Z27 and DX9 mAbs did not identify a DX9+ Z27+ population potentially corresponding to KIR3DL1*004, *019, or *054 allele products. Because KIR3DS1 and KIR3DL1*054 [32, 36] were recognized by Z27 (a weak intensity staining pattern was observed Fig. 2B), it was difficult to selectively evaluate the expression of KIR3DL1*054 at the NK-cell surface. Interestingly, the frequency of Z27+ DX9− expression was significantly higher for KIR3DS1+, 3DL1*054+ (28%, n = 7) than for KIR3DS1+, 3DL1high (13.26%, n = 19), or KIR3DS1+, KIR3DL1low (13.03%, n = 8) genotyped individuals with a p value of 0.01 and 0.03, respectively (Fig. 2C). This suggested either a cellular expression of the KIR3DL1*054 allotype, like KIR3DS1, recognized by the Z27 mAb, or a higher expression of KIR3DS1 in association with the KIR3DL1*054 allele (Fig. 2C). Different amino acid substitutions were detected between KIR3DL1*054 and KIR3DL1*004 (Fig. 2D), suggesting that KIR3DL1*054 is more similar to KIR3DS1 than to KIR3DL1. These amino acid changes, especially in the extracellular domains, could potentially interfere with Z27/DX9 epitope recognition.
KIR3DL1high NK-cell frequencies are determined by the number and nature of KIR3DL1high alleles
Taking into account the nature of KIR3DL1high alleles, we identified from all studied KIR3DL1high individuals (n = 70), a higher frequency of KIR3DL1high NK cells with the KIR3DL1*001 allele (24.3%, n = 31) than the KIR3DL1*002 allele (10.7%, n = 17, p < 0.0003), the KIR3DL1*015 allele (10%, n = 15, p = 0.01), or the KIR3DL1*008 (15.5%, n = 6, p > 0.05) (Fig. 3A). The frequency of KIR3DL1high NK cells was further compared in four groups of individuals based on the nature of the second allele: KIR3DL1high, KIR3DL1low, KIR3DL1null, and KIR3DS1. Interestingly the overall frequency of KIR3DL1high NK cells in KIR3DL1high/high individuals was about twice that found in heterozygous individuals (29.2%, n = 13 versus 12.2%, n = 57, p = 3 × 10−7) (Fig. 3B). The distribution of KIR3DL1high NK-cell frequencies was heterogeneous in KIR3DL1high/null and KIR3DL1high/3DS1 individuals, and clearly bimodal for the KIR3DL1high/null group. However, on average the KIR3DL1high NK-cell frequencies in KIR3DL1high/null and KIR3DL1high/3DS1 individuals were much lower than in KIR3DL1high/high individuals (10.8 versus 29.2%, p = 0.0015 and 12.5% versus 29.2%, p = 6.2 × 10−7, respectively). Such differences are in accordance with KIR3DL1 allelic exclusion, suggesting independent and additive generation of NK-cell subsets with the KIR3DL1 alleles being expressed in a mutually exclusive fashion. Within KIR3DL1high/high individuals (n = 13), higher KIR3DL1+ NK-cell frequencies were observed for homozygous KIR3DL1*001 individuals than for those with a single or no KIR3DL1*001 allele (Fig. 3B). A similarly higher percentage of KIR3DL1high NK cells was observed in individuals carrying KIR3DL1*001 versus other KIR3DL1high alleles in KIR3DL1high/null, KIR3DL1high/low, and KIR3DL1high/3DS1 individuals (Fig. 3B). Strikingly, the frequency of KIR3DL1high NK cells expressing the KIR3DL1*001 allele increases to around 50% with only one copy of the KIR3DL1*001 allele, notably for KIR3DL1*001,*004 genotyped individuals (Fig. 3B). Since the KIR3DL1 NK-cell frequency can be influenced by the Bw4 environment , we investigated the frequency and nature of the KIR3DL1 ligand in this French population. Overall, the frequency of HLA-B Bw4+ individuals was high (59%) and increased above 70% when both HLA-A and HLA-B Bw4+ molecules were taken into account. Fourteen out of 18 described Bw4 ligands were observed, with a predominance of A24 (27.4%) and B44 (30.2%; Supporting Information Fig. 1A). The presence of autologous HLA-A and/or HLA-B Bw4+ molecules did not affect the frequency of KIR3DL1high NK cells taking into account not only all KIR3DL1high individuals but also the number of KIR3DL1high allele copies (2X or 1X) or KIR3DL1/3DS1 allele combinations (3DL1high/low, 3DL1high/null, or 3DL1high/3DS1; Supporting Information Fig. 1B). Moreover, the higher frequency of KIR3DL1high NK cells observed with the KIR3DL1*001 allele was independent of the Bw4 ligand number (Supporting Information Fig. 1B). Thus, the heterogeneous distribution of KIR3DL1high NK-cell frequencies, especially within KIR3DL1high/null individuals, suggests the contribution of factors other than copy number and Bw4 environment. We observed that the KIR3DL1/S1 allele combinations were associated not only with KIR genotypes (Fig. 1A) but also with the presence of a particular KIR gene. For example, 21 out of 22 KIR3DL1high/3DS1 genotyped individuals displayed the KIR2DS1 gene compared with only 2 out of 12 KIR3DL1high/high individuals, none out of 23 KIR3DL1high/null and none out of 11 KIR3DL1high/low individuals (data not shown). To address whether KIR2D (KIR2DL1/2/3 and KIR2DS1/2) acquisition affects the KIR3DL1 NK-cell repertoire, we determined the phenotype of NK cells by flow cytometry using KIR3DL1-specific Z27 and KIR2DL1/2/3- and KIR2DS1/2DS2-specific 1A6 mAbs (Fig. 3C). We then analyzed the proportion of KIR3DL1high NK cells expressing KIR2D as a function of the nature of KIR3DL1high alleles (Fig. 3D, lower panel) and as a function of the KIR3DL1/S1 allele combination (Fig. 3E, lower panel). Interestingly, a proportionate fraction of KIR3DL1high KIR2D of around 50% of KIR3DL1high NK cells+ was observed regardless of the KIR3DL1high alleles and KIR3DL1/S1 allele combinations. Indeed, a positive correlation was observed mainly for KIR3DL1*001 (r = 0.956), KIR3DL1*002 (r = 0.951), and KIR3DL1*008 (r = 1) alleles with a p-value <0.05. In contrast, the Spearman's rank test showed a low correlation between the frequency of KIR2D+ KIR3DL1high+ NK cells and KIR2D+ NK-cell frequencies, with a significant p-value (<0.05) only for the KIR3DL1*001 allele (r = 0.67; Fig. 3C, upper panel) and the KIR3DL1/3DS1 allele combinations (KIR3DL1high/high, r = 0.787, KIR3DL1high/null, r = 0.775 and KIR3DL1/3DS1, r = 0.56; Fig. 3D, upper panel). These data suggest a consecutive expression of KIR2D (2DL1/2/3, 2DS1/2) on KIR3DL1high NK cells.
Bimodal distribution of KIR3DL1high NK cells in KIR3DL1high/null individuals
In KIR3DL1high/3DL1null individuals for whom a bimodal distribution of KIR3DL1+ NK cells was observed as previously mentioned, we compared the frequency and the mean fluorescence intensity of KIR3DL1high NK cells (Fig. 4A). Interestingly, among the three KIR3DL1high alleles identified (*001, *002, and *015), a positive correlation was noted between the frequency of KIR3DL1high NK cells and membrane expression level of the KIR3DL1*001 allele (r = 0.7033, p = 0.009). KIR3DL1*004 was the unique null allele associated with KIR3DL1*001, KIR3DL1*002, and KIR3DL1*015 in our cohort. Surprisingly, the KIR3DL1*001, *004 allele combination led to both a low-frequency and mean-expression level of the KIR3DL1 receptor in 5 out of 13 individuals (Fig. 4A). Representative density plots are shown in Fig. 4B, illustrating the high and low MFI of KIR3DL1+ NK cells, respectively, with high (47.3%) and low frequency (3.3%) in two representative KIR3DL1*001,*004 individuals. For comparison, a representative density plot with the expected MFI for low KIR3DL1high NK-cell frequency observed for one representative KIR3DL1*002, *004 genotyped individual is also provided in Fig. 4B. The bimodal effect observed in KIR3DL1high,null individuals was not influenced by the Bw4 ligand number (Supporting Information Fig. 1B). We explored KIR2DL1/S1, KIR2DS4, KIR2DL2, and KIR2DL3/2DS2 NK-cell subset frequencies as a function of KIR3DL1high NK-cell frequency for these KIR3DL1high,null individuals (Fig. 4C) in order to determine whether other KIR NK-cell subsets may balance the low KIR3DL1high NK-cell frequency. Individuals with high KIR3DL1high NK-cell frequencies presented a consistent KIR2DL1 NK-cell subset, however, the frequency of KIR2DL1 NK cells did not correlate with the frequency of KIR3DL1high NK cells. Interestingly, KIR2DS4 was expressed only in individuals who displayed the KIR3DL1*015 and KIR3DL1*002 alleles (Fig. 4C). Moreover, the KIR2DS2 gene, which is in linkage disequilibrium with the KIR2DL2 gene, was present in 8 out of 23 KIR3DL1high,null genotyped individuals and interestingly, KIR2DS2 was present only in KIR3DL1*001,*004 genotyped individuals who exhibited a high frequency of KIR3DL1high NK cells (data not shown). The frequency of KIR2DL2+ NK cells did not correlate with the frequency of KIR3DL1high NK cells; however, the KIR2DL3/S2 NK-cell frequency was higher in individuals who presented a high frequency of KIR3DL1high NK cells (Fig. 4C).
Because neither DX9 nor Z27 monoclonal antibodies recognize KIR3DL1 in permeabilized/fixed cells, as previously reported by Thomas et al. , any potential increased/decreased internalization of KIR3DL1 receptor could not have been evaluated in KIR3DL1*001, *004 genotyped individuals. However, Z27+ NK cells were sorted from individuals of each group and KIR3DL1 transcripts were analyzed using primers specific for KIR3DL1*001, KIR3DL1*002, or KIR3DL1*004 (Fig. 4D). We detected not only KIR3DL1*001 or KIR3DL1*002 transcripts but also KIR3DL1*004 transcripts in all studied KIR3DL1high populations. Because the amount of KIR3DL1*001 and KIR3DL1*004 transcripts might explain the different level of expression in high and low KIR3DL1 NK-cell populations, we evaluated the expression level of KIR3DL1 transcripts by real-time RT-PCR using GAPDH transcripts as an endogenous control. We observed a similar amount of both KIR3DL1*001 and KIR3DL1*004 transcripts in high and low KIR3DL1 populations from different individuals (Fig. 4E).
The nature of the KIR3DL1high allele influences KIR3DL1high NK-cell responsiveness
Among the three KIR3DL1high alleles identified (*001, *002, and *015) in KIR3DL1high/null individuals, consistent expansion of KIR3DL1high NK cells was associated only with the KIR3DL1*001 allele, as mentioned earlier. Thus, we hypothesized that these three KIR3DL1high receptors do not interact with the same affinity with the Bw4 ligand and only the KIR3DL1*001 allele product constitutes a strong inhibitory receptor capable of engaging the KIR3DL1high NK-cell subset toward cellular expansion. To test this hypothesis, we determined the degranulation potential (CD107a expression) of NK cells expressing one of these three KIR3DL1high receptors and evaluated the capacity of these three KIR3DL1high receptors to inhibit degranulation of NK cells incubated with 221-HLA-B*15:13 (Bw4 motif) target cells from 19 KIR3DL1high individuals. We observed a higher frequency of KIR3DL1*001+ NK-cell inhibition than KIR3DL1*002+ or *01502+ NK-cell inhibition upon engagement with Bw4 ligand, regardless of the frequency of KIR3DL1high NK cells or the KIR3DL1/3DS1 allele combinations (Fig. 5A). Accordingly, the percentages of KIR3DL1high NK cells inhibited by HLA-B*15:13 (Bw4 motif) were higher in KIR3DL1*001+ than in KIR3DL1*002/015+ individuals (82.3%, n = 7 versus 69.2%, n = 11, p = 0.003, Fig. 5B). Moreover, we confirmed that the spontaneous degranulation of adult KIR3DL1high NK cells against standard HLA class I-deficient 221 cells increases with the number of autologous Bw4 ligands (9.9% n = 7 versus 25.1%, n = 14, p = 0.003 between 0 and 1 Bw4 ligand and 25.1% n = 14 versus 34% n = 6, p = 0.001 between 1 and ≥2 Bw4 ligands) (Fig. 5C).
Different phenotypic profile of KIR3DL1high NK cells in early life compared with that in adulthood
Because our results suggest either a selection and specific expansion of KIR3DL1high NK-cell subsets or a deletion of KIR3DL1*001,*004 NK subsets, we investigated the KIR3DL1high NK-cell repertoire in cord blood to gain insights into KIR3DL1 NK-cell repertoire formation during development. To this end, we evaluated the frequency of this population and the expression level of the KIR3DL1high receptor taking into account the nature of the second KIR3DL1 allele (high, low, and null). Interestingly, although the KIR3DL1high NK-cell frequency was significantly lower in umbilical PBMCs (9.27%, n = 18) than adult PBMCs (19.33%, n = 68, p = 0.001) (Fig. 6A), whatever the KIR3DL1/3DS1 allele combination analyzed (Fig. 6B), the expression level was significantly higher in cord blood NK cells (umbilical MFI = 1088, n = 15 versus adult MFI = 686, n = 61, p = 3.5 × 10−7) (Fig. 6C), especially for KIR3DL1high/high (umbilical MFI = 1258, n = 4 versus adult MFI = 735, n = 11, p = 0.006) and KIR3DL1high/3DS1 allele combinations (umbilical MFI = 1268, n = 6 versus adult MFI = 689, n = 19, p = 1.5 × 10−5) reaching a MFI of 2000 (Fig. 6D). These results show that KIR3DL1high is expressed at a high intensity on the NK-cell membrane in a small proportion of NK cells during early life, probably contributing to the selection or deletion of KIR3DL1 NK-cell subsets, depending on the signaling intensity received by the cells.
The Bw4 environment impacts on KIR3DL1 NK-cell frequency in cord blood samples
Considering that a functional signal via KIR3DL1 may contribute to shaping the KIR3DL1 NK-cell repertoire, we investigated the frequency of KIR3DL1 NK cells in early life with regards to the expression or not of autologous Bw4 ligands (Fig. 6E). Interestingly, the KIR3DL1+ NK-cell frequency was significantly higher in Bw4+ cord blood (12.33%, n = 10) than in the Bw4− counterparts (4.97%, n = 7, p = 0.0003). These results suggest that an engagement toward one KIR3DL1 allele product expression during NK-cell development, possibly depending on functional interaction with Bw4 ligand, contributes to molding the KIR3DL1 NK-cell repertoire. Interestingly, although we confirmed that the spontaneous degranulation of adult KIR3DL1high NK cells increases with the number of autologous Bw4 ligands (Fig. 5C), we did not observe a significant correlation between adult KIR3DL1high NK-cell frequency and Bw4 ligand number (Supporting Information Fig. 1). This result can be partly explained by the plausible deletion of KIR3DL1 NK-cell subsets receiving excessive signals during their development, as observed for KIR3DL1high/null NK-cell subsets in the adult repertoire.
In this study, we investigated the impact of allelic KIR3DL1 polymorphism on the KIR3DL1 NK-cell repertoire. In this large cohort of healthy individuals we considered possible contributing factors, including KIR3DL1/S1 allele combinations, mean expression level, Bw4 environment, and functional potential. The French population is characterized by a large KIR3DL1 allelic polymorphism with different predominant KIR3DL1 alleles such as KIR3DL1*001, *002, *004, and *054. The frequency of KIR3DL1high (*001, *002) and KIR3DL1null (*004, *054) alleles ranges from 20 to 50% and this contrasts with the Japanese population which is characterized by (i) only one KIR3DL1high allele (*01502) overexpressed at a very high frequency (45%) compared with the next most frequent allele and (ii) the absence of the KIR3DL1*004 null allele . Surprisingly, the frequency of the null KIR3DL1*004 allele was particularly high in our cohort (29%) compared with other European or African population studies where it usually ranged from 8 to 16% . Because the Bw4 ligand could differ between populations , we have also investigated the frequency and nature of the Bw4 ligand in the French population compared with the Japanese population . The frequency of HLA-B Bw4+ was higher in the French population compared with the Japanese population (59 versus 36%) and was higher than 70% if both HLA-A and HLA-B Bw4+ were taken into account. Although the frequency of HLA-B51 (I80) was similar in both populations (13%), HLA-B44 (T80) was overrepresented in the French population (30.2 versus 7%) whereas A24 (I80) was underrepresented (27.4 versus 41%). However, in both populations, the frequency of KIR3DL1high NK cells was higher than that of the KIR3DL1low NK-cell population and the frequency of KIR3DL1high NK cells increased with gene dose. In the French population, the presence of Bw4 ligand (both HLA-A and HLA-B) or the presence of a specific HLA-Bw4 molecule such as A24, B51, B44 did not significantly increase the frequency of KIR3DL1high NK cells, while in the Japanese population, the presence of HLA-Bw4+ molecules significantly increased the frequency of NK cells expressing KIR3DL1 for donors with two high-binding 3DL1 allotypes, and the increase in frequency of KIR3DL1high NK cells was greater in Bw4 heterozygotes than in homozygotes. The fact that the Bw4 environment did not significantly affect the KIR3DL1 NK-cell frequency in our French cohort may be due to the sample size. Overall, our study confirms that French individuals are characterized by a broad KIR3DL1 allelic polymorphism in accordance with previous studies in other European populations  and that the KIR3DL1 NK-cell repertoire is mainly determined by KIR3DL1 polymorphism and KIR3DL1/3DS1 allele combinations. The extensive KIR3DL1 allelic polymorphism and diversity of KIR3DL1 ligands observed in the French population suggest a coevolution of KIR3DL1 and HLA-A, HLA-B Bw4 genes as reported in other populations .
The phenotypic study of KIR3DL1 allelic polymorphism in healthy individuals enabled us to show that all individuals hitherto described as negative for the KIR3DL1 gene actually possess an KIR3DL1*054 allele that is not detected with the current standard typing techniques, and a substantial proportion of individuals have at least one KIR3DL1 allele that is not, or only weakly, expressed on the NK-cell surface. The association of the KIR3DL1*054 allotype with KIR3DS1 in our cohort did not unable us to evaluate NK-cell surface expression of KIR3DL1*054 since both receptors seemed to be recognized similarly by the Z27 mAb . Interestingly, in our cohort, the KIR3DL1*054 allele was mainly observed in individuals with the BB KIR genotype (n = 9). While the expression and function of the KIR3DL1*004 allele is well understood,  that of the KIR3DL1*019, KIR3DL1*054 alleles remain to be elucidated. The expected mature KIR3DL1*019 protein contains the same amino acids involved in the intracellular retention of the null KIR3DL1*004 allotype, i.e. L86 and S182, and only one amino acid in the D0 domain differs between KIR3DL1*019 and KIR3DL1*004. Thus, we believe that the KIR3DL1*019 allele is also a null allele. However, different amino acid substitutions occur between KIR3DL1*054 and KIR3DL1*004 suggesting that KIR3DL1*054 is closer to KIR3DS1 than KIR3DL1. These amino acid changes in the extracellular domains could potentially interfere with Z27/DX9 epitope recognition. These results should shed light on the functional contribution of the frequent KIR3DL1*054 and KIR3DL1*004 receptors to different pathologies. Indeed, a high frequency of KIR3DL1-negative individuals has been reported in an autoimmune disease , which probably corresponded to the KIR3DL1*054 allele. Moreover, in the presence of Bw4, KIR3DL1*004 showed the most significant protection relative to all other KIR3DL1 alleles in HIV infection  and its functional expression at the cell surface has been recently demonstrated .
Our data underline the impact of KIR3DL1 allele polymorphism on its NK-cell expression pattern and on the hierarchy between all KIR3DL1 alleles, highlighting the predominance of the KIR3DL1*001 allele regardless of the presence or absence of the second allele (KIR3DL1low, KIR3DL1null, or KIR3DS1). In the KIR3DL1high,null individuals with only one copy of the KIR3DL1*001 allele, the frequency of KIR3DL1high NK cells increases up to 50% of NK cells, indicating that factors other than copy number contribute to this dominant expression. The high number of KIR3DL1*001,*004 individuals in our cohort led us to detect a bimodal distribution of KIR3DL1high NK cells: A first group with a high frequency of KIR3DL1high NK cells and high level of expression, and a second group with a low frequency of KIR3DL1high NK cells and a low level expression. Of note, we detected the same amount of KIR3DL1*001 and KIR3DL1*004 transcripts in the low and high frequency of KIR3DL1high NK cells, demonstrating that the KIR3DL1 transcript level is not linked to its protein expression at the NK-cell surface. This finding is in accordance with data demonstrating the frequent detection of KIR transcripts in T-cell clones and NK92 in the absence of the corresponding protein at the cell surface . Thus, the predominance of the KIR3DL1*001 allele in our study could not be explained by its promoter properties . Nonetheless, we cannot exclude the possibility of a differential internalization of KIR3DL1 receptor in both populations. Among other factors able to influence the frequency of KIR3DL1high NK cells, we looked at the expression of other KIR in different groups of KIR3DL1high,null individuals. Interestingly, all KIR3DL1high,null individuals with KIR3DL1*002 and KIR3DL1*015 alleles showing a low frequency and high mean level expression of the KIR3DL1 receptor expressed KIR2DS4. However, all KIR3DL1*001, *004 individuals bore only nonexpressed KIR2DS4*003 as activating KIR genes (AA genotypes) and other activating KIR genes, such as KIR2DS2, were observed in four out of eight others KIR3DL1*001, *004 genotyped individuals with a high frequency and mean level expression of KIR3DL1. Thus, since the acquisition of KIR on NK cells is sequential , it is possible that the nature of the other KIR expressed before KIR3DL1 and probably the transduction signals received by the cells, influence the posttranscriptional regulation of the KIR3DL1*001 and KIR3DL1*004 alleles. However, the bimodal distribution cannot be entirely explained by a differential expression of activating KIR.
Another explanation of the bimodal distribution could be due to a functional selection of KIR3DL1 NK cells. Indeed, we also underlined a hierarchy between KIR3DL1high alleles in their capacity to inhibit NK-cell degranulation via Bw4 engagement, and possibly to promote in vivo maintenance or expansion of peripheral NK cells carrying the corresponding allele. As such, the frequency of KIR3DL1*001+ NK-cell inhibition is higher than the frequency of KIR3DL1*002 or *015 NK-cell inhibition upon engagement with HLA-B*15:13 (Bw4 motif) expressed on transfected 221 target cells regardless of the frequency of KIR3DL1high NK cells. This last point is in agreement with a previous report indicating that highly expressed KIR3DL1*002 is a better inhibitor than poorly expressed KIR3DL1*007 . Based on recent data published by Taner et al.  and on the higher level expression of KIR3DL1 observed in cord blood, we speculate that KIR3DL1*004 can be functional, and may constitute a better signaling receptor than the KIR3DL1*002, *015, and *008 allele products, favoring its phenotypic selection during early life. Thus, for all 10 studied KIR3DL1*002/*01502, *004 genotyped individuals, we suggest that a phenotypic selection in favor of the KIR3DL1*004 receptor, which was not detected by KIR3DL1 specific Z27 and DX9 antibodies may explain the frequency of KIR3DL1high observed around 10% of NK cells. For 5 KIR3DL1*001,*004 genotyped individuals, the frequency of KIR3DL1high was low (around 10%) as was the mean fluorescent intensity, suggesting low expression of the KIR3DL1*001 receptor on NK cells. This low frequency seems to be due to a negative selection of this subset as previously described for T-lymphocyte selection. Selection and amplification of KIR3DL1high NK cells seem to be dependent on the nature of the KIR3DL1 allele and the KIR3DL1 allele combination, which is expressed with a hierarchy favoring KIR3DL1*001, followed by KIR3DL1*004, and then the other KIR3DL1high alleles (*002, *015, and *008). The engagement of two receptors such as KIR3DL1*001 and *004, thereby inducing increased signals, might lead to a negative selection of NK cells.
In this study, we did not observe any increased frequency of KIR (KIR2DL1/S1, KIR2DL2/3/S2, or KIR2DS4) NK-cell subsets balancing a low frequency of KIR3DL1high NK-cell subsets in different individuals grouped by KIR3DL1 alleles or KIR3DL1/3DS1 allele combinations (data not shown). However, we observed a constant proportion of KIR3DL1high NK cells expressing KIR2DL1/2/3, 2DS1/2 receptors, which suggests a subsequent acquisition of KIR2D by KIR3DL1high NK cells. Previous in vitro investigations of NK-cell differentiation have been performed suggesting a sequential acquisition of KIR with the expression of KIR2DL3 before KIR2DL1 . Although KIR3DL1 acquisition was observed regardless of HLA background , its acquisition in this sequential model has not been investigated. Even though it is now clear that KIR ligands contribute to the functional education of NK cells , HLA class I molecules, as KIR ligands, should partially affect the formation of the KIR NK-cell repertoire at the neonatal stage [31, 44], as observed in our study. As previously described , in our study we did not observe an impact of the Bw4 environment on the adult KIR3DL1 NK-cell repertoire. Other factors, and particularly individual immunological experience, seem to contribute to shaping the KIR NK-cell repertoire throughout life. The finding of amplified KIR+ NKG2C+ NK cells in CMV [45-47], HSV-2 , and HIV infections [49, 50] reinforces this hypothesis. In the case of KIR3DL1, it is conceivable that viral infections, particularly those involving viruses that negatively modulate HLA class I molecule expression such as CMV and HIV, trigger KIR3DL1 NK-cell subsets via KIR3DL1 or an undetermined coexpressed receptor. Further investigations in KIR3DL1high/null individuals taking into account viral status and a broader phenotype of KIR3DL1high NK cells, notably including NKG2C, should help to determine the impact of viral infection on the KIR3DL1 NK-cell repertoire. Finally, this study provides new insight into the mechanisms potentially involved in shaping the KIR3DL1 NK-cell repertoire.
Materials and methods
Cells (PBMCs, cord blood samples, and cell lines)
PBMCs were isolated from citrate-phosphate-dextrose blood, collected from healthy adult volunteers by gradient centrifugation on Ficoll-Hypaque (Lymphoprep, Axis-Shield, PoC AS, Oslo, Norway). All blood donors (n = 109) were recruited at the Blood Transfusion Center (Etablissement Français du Sang, Nantes, France) and informed consent was obtained from all individuals. Umbilical cord blood samples (n = 23) were obtained at the Nantes CHU maternity unit. Informed consent was obtained from all mothers. HLA class-I–deficient 721.221 lymphoblastoid EBV-B cells (referred to as 221 cells) and Bw4 (B*15:13) transfected 221 cells (referred to as 221-Bw4+) were used as controls to assess natural NK-cell cytotoxicity in functional assays. Cells were cultured in RPMI 1640 medium (Gibco, Paisley, Scotland, UK) containing glutamine (Gibco) and penicillin-streptomycin (Gibco) and supplemented with 10% fetal bovine serum (Gibco). Mycoplasma tests performed by PCR were negative for all cell lines.
HLA and KIR genotyping
Genomic DNA was extracted from PBMC and from cord blood samples using either a classical salting-out method  or by GenoM-6 (Qiagen, Courtaboeuf, France) using magnetic beads. Intermediate or high-resolution typing for HLA-A, HLA-B, and HLA-C was performed on all healthy donors (n = 109) and cord blood samples (n = 23) using a Sequence Based Typing kit (Abbott Molecular Park, IL, USA). All individuals were typed for the presence or absence of KIR2DL1, 2DL2, 2DL3, 2DL4, 2DL5, 3DL1, 3DL2, 3DL3, 2DS1, 2DS2, 2DS3, 2DS4, 2DS5, and 3DS1 using the KIR genotyping SSP kit from Invitrogen (Compiègne, France), lot#003, under the conditions recommended by the manufacturer. Cord blood samples were typed using a multiplex PCR-SSP method as previously described . In order to detect all KIR3DL1 alleles, especially the KIR3DL1*054 allele not amplified by the PCR-SSP Invitrogen kit lot#003, individuals negative for KIR3DL1 at the genotypic level were further typed using homemade KIR3DL1 generic primers specific for exon 3 , which amplified all KIR3DL1 alleles except KIR3DL1*027, *028, *030, *039, *042, *053, *057, *073 (all of which were amplified by the Invitrogen KIR PCR-SSP kit). For KIR3DL1 allelic typing, KIR3DL1 allele-group–specific PCR-SSP and sequencing of polymorphic exons were combined in order to resolve most of the remaining KIR3DL1 allele ambiguities arising as a result of the large KIR3DL1 allelic polymorphism described to date. For the subtyping of KIR3DL1 PCR-SSP coding sequences, primers designed to discriminate allele-group–specific polymorphisms were paired with KIR3DL1 locus-specific primers adapted from Gardiner et al. . Depending on the allele-group–specific amplification products obtained, KIR3DL1 exons 1, 2, 3, 4, 5, or 7–9 were further sequenced to resolve remaining ambiguities as previously described [17, 24]. Sequence data files were analyzed using Assign-SBT software (Conexio Genomics, Applecross, Australia) with IPD-KIR Database version 2.3.0 (August 2010).
Identification and quantification of KIR3DL1 transcripts in the KIR3DL1+ NK-cell population
For individuals genotyped as KIR3DL1*001,*004, KIR3DL1*002,*004, and KIR3DL1*002,*005, Z27+ NK cells were positively selected using KIR3DL1-specific antibody Z27 (Beckman Coulter, Immunotech) and murine IgG-coupled magnetic Dynabeads according to the manufacturer's instructions (Dynal, Oslo, Norway). Beads and KIR3DL1-specific antibody were removed using goat anti-mouse IgG antiserum (EFS) as described previously . Selected cells were stimulated using an in vitro model of NK-cell expansion . After 2 weeks of amplification, Z27+ NK-cell populations were sorted and total RNA was purified using NucleoSpin RNAXS (Macherey-Nagel). Qualitative RT-PCR was performed using a One-Step PrimeScript™ RT-PCR kit (TaKaRa, Japan) with KIR3DL1 allele-group–specific PCR-SSP primers adapted from Gardiner et al. . For relative quantification, KIR3DL1*001 and KIR3DL1*004 transcript levels were normalized to the GAPDH reference gene using the same primers and the One-Step SYBR® PrimeScript™ RT-PCR kit (TaKaRa, Japan). For each sample, qRT-PCR was performed in triplicate on a RotorGene RG6000® RT-PCR system (Corbett Research). Amplification programs were as follows: reverse transcription at 42°C for 5 min, initial denaturation at 95°C for 5 min, 5 cycles at 97°C for 20 s, 55°C (GAPDH, KIR3DL1*004) or 62°C (KIR3DL1*001) for 30 s, 72°C for 1 min, and 30 cycles at 95°C for 15 s, 57°C (GAPDH, KIR3DL1*004) or 64°C (KIR3DL1*001) for 30 s, 72°C for 1 min. A melting curve analysis was performed after each run, with one degree increments from 72 to 95°C. The relative levels of KIR3DL1 allele transcripts were compared using unpaired t-tests in different sorted NK-cell subsets.
Phenotypic analysis by flow cytometry
NK-cell surface phenotype was determined by four-color flow cytometry using the following mouse antihuman mAbs: anti-KIR3DL1/S1-PE (Z27), anti-KIR3DL1-FITC (DX9) (Beckman Coulter, Immunotech), anti-KIR2DL1/2/3/2DS1/2 (1A6) , anti-KIR2DL3/2DS2-FITC (1F12) , anti-KIR2DS4-PE (FES175), anti-KIR2DL1/S1-PE (EB6), anti-KIR2DL2/3/2DS2-PE (GL183) (Beckman Coulter, Immunotech), anti-CD3-PerCP (SK7), and anti-CD56-allophycocyanin (B159) (BD Biosciences). Cells were also stained with the corresponding isotype-matched control mAb. Data were collected using a FACSCalibur (BD Biosciences) and analyzed using Flowjo 6.2 software (TreeStar).
CD107a mobilization detected by flow cytometry
KIR3DL1high NK cells were tested for their cytolytic potential with the CD107a mobilization assay after stimulation with 221 or 221-Bw4+ (HLA-B*15:13) transfected cell lines. Preincubated NK cells with CD107-PECy5 (H4A3, BD Biosciences) were incubated with the target cells for 5 h at an effector:target ratio of 1:1, with brefeldin A (Sigma) at 10 μg/mL for the last 4 h. The cells were surface stained with Z27-PE, NKG2A-FITC and NKp46-allophycocyanin (9E2, BD Biosciences). Data were collected using a FACSCalibur (BD Biosciences) and analyzed using Flowjo 6.2 software (TreeStar).
Comparisons of KIR3DL1+ NK-cell frequencies in two different series of individuals were performed using the Student's t-test. Association between CD107a+ KIR3DL1high NK-cell frequencies and autologous Bw4 ligand number in two different series of individuals was tested using the one-way ANOVA test. p-values <0.05 were considered as statistically significant. Spearman's rank correlation coefficients were calculated and indicated only when a significant p-value was obtained (p < 0.05).
We thank Dr. Marc Bonneville (UMRS892, Nantes, France) for advice, Dr. Damian Goodridge (Conexio Genomics, Western Australia) for his help in building and optimizing the KIR3DL1 libraries for use in Assign software and Britt House (Nantes, France) for editing the manuscript. This work was financially supported by the EFS Pays de la Loire and by grants from the “Nantes University”, “Association Recherche et Transfusion”, “IRGHET”, “Agence de la BioMédecine”, “Etablissement Français du Sang 2011–06”, and NAGMO. ZD is a PhD student supported by a CIFRE grant (N°447/2011) and PR is a PhD student supported by an EFS P.L/Region Pays de la Loire grant.
Conflict of interest
The authors declare no financial or commercial conflict of interest.