Centromeric KIR AA Individuals Harbor Particular KIR Alleles Conferring Beneficial NK Cell Features with Implications in Haplo-Identical Hematopoietic Stem Cell Transplantation

Simple Summary We have recently shown a broad disparity of Natural Killer (NK) cell responses against leukemia, highlighting good and bad responders resting on the Killer cell Immunoglobulin-like Receptors (KIR) and HLA genetics. In this study, we deeply investigated KIR2DL NK cell repertoire in combining high-resolution KIR allele typing and multicolor flow cytometry from a cohort of 108 blood donors. Our data suggest that centromeric (cen) AA individuals display more efficient KIR2DL alleles (L1*003 and L3*001) to mount a consistent frequency of KIR2DL+ NK cells and to confer an effective NK cell responsiveness. The transposition of our in vitro observations in T-replete haplo-identical Hematopoietic Stem Cell Transplantation (HSCT) context led us to observe that cenAA HSC grafts limit significantly the incidence of relapse in patients with myeloid diseases after T-replete haplo-identical HSCT. As NK cells are crucial in HSCT reconstitution, one could expect that the consideration of KIR2DL1/2/3 allelic polymorphism could help to refine scores used for HSC donor selection. Abstract We have recently shown a broad disparity of Natural Killer (NK) cell responses against leukemia highlighting good and bad responders resting on the Killer cell Immunoglobulin-like Receptors (KIR) and HLA genetics. In this study, we deeply studied KIR2D allele expression, HLA-C recognition and functional effect on NK cells in 108 blood donors in combining high-resolution KIR allele typing and multicolor flow cytometry. The KIR2DL1*003 allotype is associated with centromeric (cen) AA motif and confers the highest NK cell frequency, expression level and strength of KIR/HLA-C interactions compared to the KIR2DL1*002 and KIR2DL1*004 allotypes respectively associated with cenAB and BB motifs. KIR2DL2*001 and *003 allotypes negatively affect the frequency of KIR2DL1+ and KIR2DL3+ NK cells. Altogether, our data suggest that cenAA individuals display more efficient KIR2DL alleles (L1*003 and L3*001) to mount a consistent frequency of KIR2DL+ NK cells and to confer an effective NK cell responsiveness. The transposition of our in vitro observations in the T-replete haplo-identical HSCT context led us to observe that cenAA HSC grafts limit significantly the incidence of relapse in patients with myeloid diseases after T-replete haplo-identical HSCT. As NK cells are crucial in HSCT reconstitution, one could expect that the consideration of KIR2DL1/2/3 allelic polymorphism could help to refine scores used for HSC donor selection.


Introduction
Natural Killer (NK) cells are granular lymphocytes and form part of the innate immune system. They sense the absence of or decreased HLA class I expression on tumoral or virally infected cells and allogenic cells. They are particularly important in Hematopoietic Stem Cell Transplantation (HSCT), being the first cytotoxic lymphocytes to appear during immune reconstitution before T cell recovery [1], and they are engaged in the beneficial Graft-versus Leukemia (GvL) effect [2,3]. This missing-self recognition by NK cells [4] is mediated through different receptors specific for HLA class I molecules [5]. Among these receptors, Killer cell Immunoglobulin-like Receptors (KIR) play a major role in the functional education of NK cells [6] and the modulation of NK cell functions [7,8]. The absence of inhibitory KIR engagement with its cognate ligand results in triggering KIR + NK cell functions.
The structure and the function of the NK cell repertoire depends on the clonal expression of different KIR combinations defined by the number and the nature of KIR genes and alleles, the HLA class I environment and the immunological history, notably cytomegalovirus (CMV) infection [8,9].
In humans, 15 KIR genes, located on the chromosome 19q13.4, have been identified (Figure 1a) [9]. The number and the nature of KIR genes differ according to individuals allowing the discrimination of two defined KIR haplotypes [10]. The A KIR haplotype is defined by a fixed set of nine KIR genes, including KIR2DS4 as the only activating KIR gene. In contrast, B haplotypes are more diverse, with a variable number of KIR genes, and are characterized by the presence of more than one activating KIR gene and the absence of the KIR2DS4 gene ( Figure 1a) [10].
The arrangement of KIR genes and the high sequence similarity facilitate gene gain and loss. Thus, all KIR loci are subject to copy number variation (CNV), particularly in B haplotypes [11]. For instance, the KIR2DL1 gene, present on both A and B haplotypes (Figure 1a), is mainly observed in the form of one copy per haplotype and less than 20% of KIR haplotypes do not display the KIR2DL1 gene [11]. Based on the combination of A and B KIR haplotypes, 660 KIR genotypes were described worldwide (http://www.allelefrequencies.net/). Depending on KIR gene content, centromeric and telomeric KIR motifs were also defined ( Figure 1b) [12]. KIR gene length and exon/intron organization vary as illustrated for KIR2DL1/2/3 genes (Figure 1c). Inhibitory KIRs display immuno-receptor tyrosine-based motifs (ITIM) in their cytoplasmic tails (Figure 1d). In contrast, activating KIRs are coupled to adaptor DAP12 protein that contains immunoreceptor tyrosine-based activating motifs (ITAMs). In particular, KIR2DL1 recognizes exclusively HLA-Cw molecules belonging to the group C2 epitope (Lys80) [13] whereas KIR2DL2/3 recognizes HLA-Cw molecules belonging to the C1 group epitope (Asn80) and some HLA-Cw molecules of C2 group (Figure 1d) [14,15].  19 in the Leucocyte Region Complex (LRC), which also contains genes encoding DAP adaptor protein and other Natural Killer (NK) cell receptors. The A KIR haplotype is defined by a fixed set of nine genes with only 2DS4 as activating genes, whereas B haplotypes are more diverse and characterized by the presence of more than one activating KIR gene and the absence of 2DS4 (a). Different KIR centromeric (cen) and telomeric (tel) KIR motifs are defined depending on the presence or absence of KIR2DL3/L2/S2 and KIR2DS1/S4/3DL1/S1 genes respectively (b). KIR2DL1/L2/L3 genes have eight exons, as well as a pseudoexon 3 coding for two Ig-like domains and a long intracytoplasmic tail (c). KIR2DL1/L2/L3/S1/S2 receptors interact with specific HLA-C molecules divided into C1 or C2 group (d).
The arrangement of KIR genes and the high sequence similarity facilitate gene gain and loss. Thus, all KIR loci are subject to copy number variation (CNV), particularly in B haplotypes [11]. For instance, the KIR2DL1 gene, present on both A and B haplotypes (Figure 1a), is mainly observed in the form of one copy per haplotype and less than 20% of KIR haplotypes do not display the KIR2DL1 gene [11]. Based on the combination of A and B KIR haplotypes, 660 KIR genotypes were described worldwide (http://www.allelefrequencies.net/). Depending on KIR gene content, centromeric and telomeric KIR motifs were also defined ( Figure 1b) [12]. KIR gene length and exon/intron organization vary as illustrated for KIR2DL1/2/3 genes (Figure 1c). Inhibitory KIRs display immunoreceptor tyrosine-based motifs (ITIM) in their cytoplasmic tails (Figure 1d). In contrast, activating KIRs are coupled to adaptor DAP12 protein that contains immunoreceptor tyrosine-based activating motifs (ITAMs). In particular, KIR2DL1 recognizes exclusively HLA-Cw molecules belonging to the group C2 epitope (Lys80) [13] whereas KIR2DL2/3 recognizes HLA-Cw molecules belonging to the C1 group epitope (Asn80) and some HLA-Cw molecules of C2 group (Figure 1d) [14,15].
KIR genes exhibit an allelic diversity, with 1110 KIR alleles currently referenced (https://www.ebi.ac.uk/ipd/kir/). Of note, this allelic polymorphism is more important for the inhibitory rather than for the activating KIR genes. Although some studies have underlined the effect of KIR allelic polymorphism on the phenotype and the function of NK cells as reported for KIR3DL1 [16,17] and for KIR2DL1 [18][19][20], the absence of high-resolution methods limited investigations on KIR + NK cell repertoire. Clinical consequences of KIR allelic polymorphism on KIR diseases were Figure 1. Localization, organization of Killer cell Immunoglobulin-like Receptors (KIR) genes and HLA-C specificity of KIR2DL interactions. KIR genes are located on chromosome 19 in the Leucocyte Region Complex (LRC), which also contains genes encoding DAP adaptor protein and other Natural Killer (NK) cell receptors. The A KIR haplotype is defined by a fixed set of nine genes with only 2DS4 as activating genes, whereas B haplotypes are more diverse and characterized by the presence of more than one activating KIR gene and the absence of 2DS4 (a). Different KIR centromeric (cen) and telomeric (tel) KIR motifs are defined depending on the presence or absence of KIR2DL3/L2/S2 and KIR2DS1/S4/3DL1/S1 genes respectively (b). KIR2DL1/L2/L3 genes have eight exons, as well as a pseudoexon 3 coding for two Ig-like domains and a long intracytoplasmic tail (c). KIR2DL1/L2/L3/S1/S2 receptors interact with specific HLA-C molecules divided into C1 or C2 group (d).
KIR genes exhibit an allelic diversity, with 1110 KIR alleles currently referenced (https://www. ebi.ac.uk/ipd/kir/). Of note, this allelic polymorphism is more important for the inhibitory rather than for the activating KIR genes. Although some studies have underlined the effect of KIR allelic polymorphism on the phenotype and the function of NK cells as reported for KIR3DL1 [16,17] and for KIR2DL1 [18][19][20], the absence of high-resolution methods limited investigations on KIR + NK cell repertoire. Clinical consequences of KIR allelic polymorphism on KIR diseases were identified for a few KIR genes such as KIR2DL1 for preeclampsia [21], and KIR2DL1 [22], KIR3DL1 [23] for HSCT outcome. Besides KIR allelic polymorphism, HSC donors with a B KIR gene motif are preferentially chosen to decrease relapse incidence after unrelated HLA identical HSCT [12,24], supporting a role of activating KIR or KIR2DL2. This protection against relapse in patients with acute myeloid leukemia (AML) is observed after myeloablative (MAC) and reduced intensity conditioning (RIC) regimens [12,24]. In the context of T-replete HLA haplo-identical HSCT using post-transplant cyclophosphamide (haplo-PTCY), some studies have reported the impact of donor KIR genotypes and/or KIR/HLA mismatches on patient outcome after haplo-PTCY [25][26][27][28][29][30]. In particular, we have shown that KIR2DL/HLA incompatibilities are beneficial to improve patient's outcome after haplo-PTCY [30]. In haplo-PTCY for which multiple options of HSC donors are feasible, the question arises of whether genetic markers as KIR help to select the best one to improve efficient immune reconstitution commingled with beneficial GvL effect. To address this question, we deeply investigated the influence of KIR2DL1/2/3 allelic polymorphism on Cancers 2020, 12, 3595 4 of 18 the phenotypic and functional structuration of the NK cell repertoire from 108 healthy individuals in combining high-resolution KIR allele typing by Next-Generation Sequencing (NGS) [31], high-resolution flow cytometry using KIR specific mAbs [32] and in silico KIR/HLA-C modelling. In parallel, our in vitro results were compared to ex vivo observations in 81 T-replete haplo-identical HSCT patients treated with PTCY in order to evaluate the predictive dimension of KIR to improve HSC donor selection.

Results
2.1. Predominant KIR2DL1*003 and KIR2DL3*001 Alleles Are Frequently Identified as Unique Allele and Are Associated with Centromeric AA Motifs In this study, we focused our investigations on KIR2DL1/2/3 allelic diversity as their gene frequencies are close to 100%, as well as because they are the most engaged inhibitory receptors with HLA-C ligands. On 97 KIR2DL1 + genotyped healthy individuals, 5 KIR2DL1 alleles were identified. The most frequent KIR2DL1 allele was KIR2DL1*003 (39.7%), followed by *002 (33.3%), *004 (20.7%), *001 (2.5%) and *007 (2.5%). Interestingly, KIR2DL1*001 and *004 were more frequently associated with another KIR2DL1 allele than found alone ( Figure 2a). In contrast, KIR2DL1*003 was predominantly found as unique (Figure 2a). and/or KIR/HLA mismatches on patient outcome after haplo-PTCY [25][26][27][28][29][30]. In particular, we have shown that KIR2DL/HLA incompatibilities are beneficial to improve patient's outcome after haplo-PTCY [30]. In haplo-PTCY for which multiple options of HSC donors are feasible, the question arises of whether genetic markers as KIR help to select the best one to improve efficient immune reconstitution commingled with beneficial GvL effect. To address this question, we deeply investigated the influence of KIR2DL1/2/3 allelic polymorphism on the phenotypic and functional structuration of the NK cell repertoire from 108 healthy individuals in combining high-resolution KIR allele typing by Next-Generation Sequencing (NGS) [31], high-resolution flow cytometry using KIR specific mAbs [32] and in silico KIR/HLA-C modelling. In parallel, our in vitro results were compared to ex vivo observations in 81 T-replete haplo-identical HSCT patients treated with PTCY in order to evaluate the predictive dimension of KIR to improve HSC donor selection.
were very outlying in cenAB individuals suggesting that other parameters may modulate this response. KIR2DL2*001 + NK cells observed mainly in cenAB individuals showed the lowest degranulation potential compared to L3*001 + , L3*002 + and L3*005 + NK cells in C1 + individuals ( Figure  5f). No difference can be highlighted between KIR2DL3 allotypes due to the limited size number of L3*005 and L3*002 individuals. Of note, KIR2DL2*003 + NK cell degranulation frequencies associated with cenBB motifs were higher even though disparate.
We assessed the inhibition of KIR2DL + NK cell degranulation against a panel of 3 C1 and four C2 transfected 221 cell lines focusing on main KIR2DL1/2/3 allotypes. The C2 specificity of KIR2DL1 + NK cells was confirmed with nonetheless a hierarchy of recognition toward different HLA-C molecules belonging to the C2 group (data not shown [34]). In contrast, 2DL2 and 2DL3 did not exclusively recognize C1 ligands since a strong inhibition of 2DL2 + and 2DL3 + NK was observed with the 221-HLA-C*04:01 cell line, being C2 + target (data not shown [34]) as we previously reported [15].

Beneficial Impact of cenAA HSC Donors on Relapse Incidence after T-Replete Haplo-Identical HSCT in Myeloid Diseases
Our previous observations support that cenAA L1*003 and L3*001 allotypes are associated with a higher frequency of KIR2DL + NK cells and better responsiveness. We further determine the centromeric KIR gene motifs of HSC donors. On 81 included haplo-PTCY patients, 39 cenAA and 42 cenAB.BB (cenB + ) donors were identified (Table S2). Haplo-PTCY performed from cenAA or cenB + donors were identical for the clinical characteristics including diseases, status at treatment, disease risk index and conditioning (Table S2). The relapse incidence between haplo-PTCY performed from cenAA vs. cenB + donors did not differ when all patients were included (Figure 7a). However, relapse incidence was decreased with cenAA donors were compared to cenB + HSC donors for patients with myeloid diseases (Figure 7b) but not for patients with lymphoid diseases (Figure 7c).
We assessed the inhibition of KIR2DL + NK cell degranulation against a panel of 3 C1 and four C2 transfected 221 cell lines focusing on main KIR2DL1/2/3 allotypes. The C2 specificity of KIR2DL1 + NK cells was confirmed with nonetheless a hierarchy of recognition toward different HLA-C molecules belonging to the C2 group (data not shown [34]). In contrast, 2DL2 and 2DL3 did not exclusively recognize C1 ligands since a strong inhibition of 2DL2 + and 2DL3 + NK was observed with the 221-HLA-C*04:01 cell line, being C2 + target (data not shown [34]) as we previously reported [15].

Beneficial Impact of cenAA HSC Donors on Relapse Incidence after T-Replete Haplo-Identical HSCT in Myeloid Diseases
Our previous observations support that cenAA L1*003 and L3*001 allotypes are associated with a higher frequency of KIR2DL + NK cells and better responsiveness. We further determine the centromeric KIR gene motifs of HSC donors. On 81 included haplo-PTCY patients, 39 cenAA and 42 cenAB.BB (cenB + ) donors were identified (Table S2). Haplo-PTCY performed from cenAA or cenB + donors were identical for the clinical characteristics including diseases, status at treatment, disease risk index and conditioning (Table S2). The relapse incidence between haplo-PTCY performed from cenAA vs. cenB + donors did not differ when all patients were included (Figure 7a). However, relapse incidence was decreased with cenAA donors were compared to cenB + HSC donors for patients with myeloid diseases (Figure 7b) but not for patients with lymphoid diseases (Figure 7c).  To assert that cenAA HSC grafts significantly limit relapse incidence in patients with myeloid malignancies after haplo-PTCY, we performed univariate and multivariate analyses including as variables age, recipient gender, status at treatment, disease risk index (DRI), conditioning and donor HSC cenAA motif. Univariate analysis identified DRI (high/very high vs. intermediate) as the most significant factor predicting relapse (HR = 4.07 [95%CI 1.87-8.89], p = 0.0004). There was a trend for a lesser relapse rate in patients with myeloid diseases grafted with cenAA HSC donors (HR = 0.50 [95%CI 0.23-1.09], p = 0.08). Age, gender, status at treatment and conditioning were not significantly associated with relapse (data not shown [34]). Multivariate analysis confirmed that donor HSC cenAA motifs were significantly associated with decreased relapse in contrast to DRI that is associated with increased relapse in patients with myeloid diseases after haplo-PTCY (Table 1). Overall, these results sustain a beneficial effect of cenAA donors on relapse incidence after haplo-PTCY only in myeloid diseases, arguing for a better GvL effect driven by NK cells.

Discussion
We confirmed a limited number of alleles for each KIR2DL receptor with closed frequencies reported in other European cohorts [35] and an association with KIR A (L*001, *002 and *003 alleles) and B (L1*004) haplotypes [21,[36][37][38]. For the first time, we highlighted the dominance of L1*003 and L3*001 alleles in cenAA individuals. Moreover, we showed that L1*002, L3*002 and L2*001 alleles were predominant in cenAB whereas L1*004 or no 2DL1 and L2*003 alleles were predominant in cenBB individuals. We refined some LD [39,40], taking into account 2DL1/2/3/S1/S2 allele combinations encountered in both cen and tel motifs, although haplotype family segregation was lacking in our study. Thus, the clustering taking into account only cenA and B motifs is probably not completely accurate. KIR2DL1/2/3 allotypes confer different phenotypic and functional characteristics to NK cells based on KIR cen motifs. CenAA individuals display mainly the L1*003 allotype with highest frequency, expression level [19][20][21]38] and strength KIR/HLA-C interactions compared to the L1*002 allotype in cenAB and to the L1*004 allotype in cenBB individuals. In this study, we did not determine the CNV for KIR2DL1. However, although the frequency of NK cells expressing a given KIR correlates with the CNV of that gene [38], the coexpression of multiple copies is infrequent [41]. Thanks to the combination of KIR specific 1A6 and 8C11 mAbs [32], we discriminate for the first time the main KIR2DL1 allotypes. We demonstrated that KIR2DL1 + NK cell degranulation was higher with the L1*003 allotype, usually associated with cenAA motifs.
The stronger recognition of C2 ligands by KIR2DL1 + NK cells was observed for the KIR2DL1*003 allotype as previously observed [18]. In contrast to other studies pooling KIR2DL1*002 and *003 allotypes with R245 residue [18,20], we were able to discriminate both KIR2DL1 allotypes showing that the L1*002 allotype (R245) interacts with all the C2 + targets similar with the L1*004 allotype (C245). The modelling of the studied KIR2DL1 allotypes associated with the HLA-Cw4 molecule [13] suggests that the difference in functionality between the allotypes is not due to the polymorphic positions located on the domains binding of the HLA-C molecule. However, we cannot exclude an influence of loaded peptides, which are probably different in our in vitro cellular model and in in silico modelling. Although a minimal role of peptide has been deduced from the crystal structure of KIR2DL1/HLA-Cw4 complex [13], studies with synthetic peptide analogs have showed that substitution of Lys8 in the peptide with acid residue results in KIR binding loss [42,43].
The KIR2DL1*004 allele associated with cenBB motifs confers a lower degranulation potential to NK cells. It negatively affects the frequency of KIR2DL1 + and 2DL3 + NK cells and the level expression of L1*003 and *002 allotypes. In contrast, it positively affects the level expression of L3*001 allotype when associated with cenAB motifs. In absence of KIR2DL2 specific mAb, it is difficult to grasp KIR2DL2 + NK cell frequency and the expression level of KIR2DL2. We observed that KIR2DL2 + NK cell degranulation was more heterogeneous than L3 + NK cells without identifying link with specific genetic parameters. Educated KIR2DL2*001 + NK cells harbored the worst degranulation compared to all KIR2DL2/3 allotypes. The L2*001 allele is often observed in cenAB individuals in combination with L3*001 or L3*002 alleles whereas L2*003 is mainly observed in cenBB donors without KIR2DL3. This observation suggests that KIR2DL3 could negatively regulate the function of the L2*001 allotype in cenAB individuals.
Heterogeneous levels of inhibition of KIR2DL2/3 + NK cell degranulation was observed mainly against less stringent ligands as HLA-C*08:02 and -C*02:02, -C*06:02 showing a better inhibition even not significant of 2DL2 than L3 allotypes. HLA-C allelic polymorphism could affect the KIR2DL2/3/HLA-C affinity and the functionality of NK cells, both from a conformational and on the level of expression of this ligand [44,45]. The peptide presented by HLA-C molecules could also affect the KIR2DL2/3-ligand affinity [46]. In contrast to KIR2DL1, the peptide modulates the binding of KIR2DL2/3 to the HLA-Cw3 through direct interactions [14,42]. Studies have also shown that KIR2DL2/3 have a higher affinity with the HLA-C*03:04 allotype loaded with Hepatitis Chronic Virus (HCV)-derived peptides [46]. Altogether, we suggest that cenAA individuals display more efficient KIR2DL alleles (L1*003 and L3*001) to mount a consistent frequency of KIR2DL + NK cells and to confer an effective spontaneous degranulation of NK cells. Nonetheless, further investigations on a broader cohort would be necessary to include rare KIR2DL alleles not documented in our study. Moreover, our functional results raise additional questions concerning the influence of HLA-C polymorphism. Indeed, we observed a broad spectrum of HLA-C recognition by KIR2DL independently of C1/C2 classification, suggesting an impact of HLA-C polymorphism on the phenotypic and functional structuration of the NK cell repertoire. Finally, we did not include peptide influence that seems to modulate KIR2DL affinity.
Some studies have reported either a beneficial [25,28] or no impact [26] of donor KIR B genotypes on relapse incidence after haplo-PTCY. However, the impact of donor KIR centromeric motifs on clinical outcome was not investigated so far after T-replete haplo-PTCY. Here, we report with multivariate analysis that myeloid patients grafted with HSC donors harboring a KIR cenAA motif have a lower incidence of relapse compared to cenB + donors after haplo-PTCY. In addition, multivariate analysis showed that DRI was the most significant factor affecting relapse incidence in myeloid patients after haplo-PTCY, as previously reported [47,48]. By contrast, no protective effect of donor B + genotype was shown on relapse incidence (data not shown [34]). Heterogeneities concerning the proportion of AML patients, conditioning regimen and stem cell source between published studies [25,26,28] and that reported here could explain these discordances. The protective effect of cenAA donors we report here was observed whatever the immunosuppressive regimen and only in myeloid patients, including 59% of AML patients. This is in agreement with our recent work showing that KIR + NK cell subsets are preferentially engaged against AML [49]. Overall, these clinical data suggest that the GvL effect could be driven by KIR2DL1 or 2DL3 specific of cenAA motifs after haplo-PTCY.

Healthy Individuals
One hundred and eight blood donors were recruited at the Blood Transfusion Center (Etablissement Français du Sang, Nantes, France) and informed consent was given by all donors. Preparation and conservation of these biocollections have been declared to French Research's Minister (DC-2014-2340) and has received approval from the IRB (2015-DC-1).

Cohort of T-Replete Haplo-Identical HSCT Patients
This retrospective study has included 81 adult patients with various hematological malignancies who received a T cell-replete haploidentical HSCT with post-transplant cyclophosphamide (haplo-PTCY) in the Hematology Department of Nantes University Hospital. Conditioning regimens consisted of a Baltimore-based RIC regimen with fludarabine (n = 26) [50] or clofarabine (Clo-Baltimore, n = 27) [51] or a CloB2A1 regimen with clofarabine 30 mg/m 2 /day (d), d-6 to -2, busulfan 3.4 mg/kg d-3 and d-2, and ATG 2.5 mg/kg d-1 (n = 28) [52]. The source of graft was peripheral blood stem cell (PBSC) for all cases. The study complies with the Declaration of Helsinki. All patients provided informed consent for collecting their own data from the PROMISE database of the European Bone Marrow Transplantation (EBMT). This study was approved by the Ethics Review Board of the Nantes University Hospital and all patients and HSC donors provided informed consent. The clinical outcome and immune reconstitution of some patients have been previously reported [30,53] and have been updated in October 2020 for this study. The primary endpoint was relapse incidence.

HLA Class I and KIR Genotyping
HLA-A, -B, and -C typing was carried out by Next-Generation-Sequencing (NGS) using Omixon Holotype HLA ® (Omixon, Budapest, Hungary). Generic KIR typing was performed on all individuals and HSC donors using a KIR multiplex PCR-SSP method [55]. Centromeric and telomeric KIR motifs were defined taking into account KIR2DL2/3/S2 and KIR3DL1/S1/2DS1/2DS4 genes respectively as reported [12]. KIR2DL1/2/3/S1/S2 alleles were assigned on all healthy individuals by NGS [31]. KIR genes were firstly captured by Long Range PCR using five intergenic KIR primers according the LR-PCR protocol already described [31]. Qubit dsDNA high sensitivity Assay kit (Life technologies, Villebon sur Yvette, France) was used to quantify the starting DNA library on a Qubit ® fluorometer (Life technologies). The KIR library preparation was performed using the NGSgo GENDX kit (Bedia Genomics, Chavenay, France) according to the manufacturer's instructions. The final denatured library was subsequently sequenced by using a MiSeq sequencer (HLA laboratory, EFS Nantes, France) with 500 cycles v2 kits which generated 250-base paired-end sequence reads. The quality of raw data sequences was monitored by using the Sequencing Analysis Viewer (SAV) Illumina software. KIR2DL1/2/3/S1/S2 allele assignment was performed by using the Profiler software developed by Dr M. Alizadeh (Research Laboratory, Blood Bank, Rennes, France) [31]. An updated KIR allele library (v2.9) was implemented into the Profiler software version 2.24.

Statistical Analysis
Categorical data were analyzed by Chi-square test and univariate comparisons were performed by the Student t test. Statistical differences in KIR2DL + NK cell frequencies between individuals having different KIR2DL1/2/3 alleles were analyzed with unpaired t tests or one-way ANOVA test for multiple comparisons using the GraphPad Prism v6.0 software (San Diego, CA, USA). Clinical and demographic variables for patients (i.e., age, gender, status at treatment, disease risk index, conditioning and donor centromeric AA KIR motif) were evaluated for their impact on relapse incidence in univariate and multivariate analyses using LogRank test and Cox proportional hazards models adjusted for significant clinical factors. Multivariate analysis was performed including only variables having a p-value less than 0.20. Univariate and multivariate analyses were performed with the Medcalc (Ostend, Belgium) software. p values < 0.05 were considered statistically significant.

Conclusions
Altogether, our data suggest that cenAA individuals display more efficient KIR2DL alleles (L1*003 and L3*001) to mount a consistent frequency of KIR2DL + NK cells and to confer an effective NK cell responsiveness. The transposition of our in vitro observations in T-replete haplo-identical HSCT context led us to observe that cenAA HSC grafts limit significantly the incidence of relapse in patients with myeloid diseases after T-replete haplo-identical HSCT. Nevertheless, our conclusions have to be taken with caution and to be confirmed from a larger cohort of haplo-identical HSCT donor/recipient pairs. NK cell characteristics are crucial in HSCT, one could expect that the consideration of KIR2DL1/2/3 allelic polymorphism could help to refine scores used for HSC donor selection, and to evaluate its influence on HSCT outcome.

Supplementary Materials:
The following are available online at http://www.mdpi.com/2072-6694/12/12/3595/s1, Figure S1: Structural model of KIR2DL1 allotypes complexed with HLA-Cw4 molecule, Table S1: Numerous KIR2DL1/2/3/S1/S2 allele combinations encountered in a cohort of French blood donors (n = 108) depending on KIR gene motifs and HLA-Cw environment. Table S2: Characteristics of patients.  San Diego), and then manually optimized. The X-ray crystallographic structure of the KIR2DL1 bound to HLA-Cw4 (PDB code 1IM9) was used to obtain the structural models of the KIR2DL1 allotypes/HLA-Cw4 complexes. The structure was first prepared by adding hydrogen atoms, removing water molecules, and inserting the missing loop region using the Prepare Protein tool within DS. Non-bound KIR2DL1*002 allotype structure was extracted from the prepared structure complex (100% identity) and used to obtain the structural model of the two other non-bound KIRD2DL1 allotypes (*003 and *004) using the "Build Mutant" tool of DS with secondary restraints and a distance cut-off of 5 Å centered on the mutate residues for including neighboring residues in optimization. The generated models were checked using Verify_3D (https://servicesn.mbi.ucla.edu/Verify3D) and Procheck (https://servicesn.mbi.ucla.edu/PROCHECK) programs and the structures with the best stereochemical and folding qualities were retained. The structural models of the KIR2DL1 allotypes have then undergone a series of energy minimizations with CHARMm force field implemented under DS in a two-step procedure to relax progressively the structures (1: all heavy atoms fixed; 2: heavy atoms of the backbone fixed) using steepest descent steps with convergence obtained at 0.001 and 0.01 rmsg respectively. Finally, the structural models of the three KIR2DL1 allotypes bound to the HLA-Cw4 molecule were obtained by superimposing the three non-bound KIR2DL1 allotype structural models into the prepared KIR2DL1/HLA-Cw4 complex structure. A last minimization step (steepest descent with convergence obtained at 0.01 rmsg and fixed constraints for the backbone atoms and for the heavy atoms of all residues outside the KIR2DL1/HLA-Cw4 interface) was then carried out to avoid atom clashes and to optimize interactions in the interface.