Non-Expressed Donor KIR3DL1 Alleles May Represent a Risk Factor for Relapse after T-Replete Haploidentical Hematopoietic Stem Cell Transplantation

Simple Summary Natural killer (NK) cells are key cytotoxic effectors against leukemic cells. The polymorphism of killer cell immunoglobulin-like receptor (KIR) genes plays a crucial role in the NK cell repertoire. In particular, different levels of KIR3DL1 expression on the NK cell surface are described, discriminating non-expressed vs. expressed allotypes depending on the KIR3DL1 alleles. KIR3DL1 allelic polymorphism after T-replete haploidentical hematopoietic stem cell transplantation (hHSCT) has not yet been investigated. In this study, we first assessed the extent of non-expressed versus expressed KIR3DL1 allotypes in a cohort of healthy blood donors and then evaluated their clinical impact on relapse incidence after hHSCT. Overall, we would expect that taking KIR3DL1 allelic polymorphism into consideration could help to refine the scores used for HSC donor selection. Abstract KIR3DL1 alleles are expressed at different levels on the natural killer (NK) cell surface. In particular, the non-expressed KIR3DL1*004 allele appears to be common in Caucasian populations. However, the overall distribution of non-expressed KIR3DL1 alleles and their clinical relevance after T-replete haploidentical hematopoietic stem cell transplantation (hHSCT) with post-transplant cyclophosphamide remain poorly documented in European populations. In a cohort of French blood donors (N = 278), we compared the distribution of expressed and non-expressed KIR3DL1 alleles using next-generation sequencing (NGS) technology combined with multi-color flow cytometry. We confirmed the predominance of the non-expressed KIR3DL1*004 allele. Using allele-specific constructs, the phenotype and function of the uncommon KIR3DL1*019 allotype were characterized using the Jurkat T cell line and NKL transfectants. Although poorly expressed on the NK cell surface, KIR3DL1*019 is retained within NK cells, where it induces missing self-recognition of the Bw4 epitope. Transposing our in vitro observations to a cohort of hHSCT patients (N = 186) led us to observe that non-expressed KIR3DL1 HSC grafts increased the incidence of relapse in patients with myeloid diseases. Non-expressed KIR3DL1 alleles could, therefore, influence the outcome of hHSCT.


Introduction
Killer cell immunoglobulin-like receptors (KIRs) play a crucial role in the structure of the natural killer (NK) cell receptor repertoire and the education of NK cells through interactions with self-HLA class I molecules [1,2]. KIRs comprise a family of inhibitory (2DL, 3DL) and activating (2DS, 3DS) receptors clonally expressed on NK cells and some T cell subsets [3]. All HLA-Cw variants function as specific ligands for KIR2DL1/2/3, whereas only HLA-A and HLA-B molecules represent the Bw4 epitope function as ligands for KIR3DL1 [4]. KIR genes exhibit a specific organization, including variations in their content, haplotypes, centromeric (cen) and telomeric (tel) motifs, and allelic polymorphisms [5]. Among the KIRs, the inhibitory KIR3DL1 and its activating KIR3DS1 counterpart are segregated as alleles of the same gene, and there is the possibility that individuals can have multiple copies of KIR3DL1/S1 [6]; thus, they are among the most intriguing KIRs [7]. While the KIR3DL1 gene is present on the A KIR haplotype at a frequency close to 95%, KIR3DS1 is, at present, only on the B KIR haplotype (40%). Thus, the inheritance of KIR3DL1 or KIR3DS1 defines different tel motifs [8]. In addition, an extensive allelic polymorphism with 189 KIR3DL1 and 91 KIR3DS1 alleles has been described and reported in the last IPD KIR database.
We previously reported that the KIR3DL1/S1 gene has an impact on HSCT outcomes [33,34] and that the nature of both KIR3DL1 alleles and the KIR3DL1/S1 allele combination is involved in modulating the repertoire of KIR3DL1 + NK cells [35]. Notably, we reported a high proportion of individuals who had a null KIR3DL1 allele, such as KIR3DL1*004, L1*019, and L1*054 [35], as well as the induction of KIR3DS1 expression on NK cells by various stimuli in KIR3DS1 + /KIR3DL1 null individuals [36]. The roles of 3DL1*019, first identified in Caucasian individuals [37], and 3DL1*054 [22], which are less frequent than KIR3DL1*004, remain unknown. In this study of a cohort of volunteer blood donors, the distribution of non-expressed vs. expressed KIR3DL1 alleles was assessed using a high-resolution next-generation sequencing (NGS) technology that we developed [38], which was used to characterize the phenotype and Bw4 recognition of the KIR3DL1*019 allotype using specific KIR3DL1 constructs and dedicated KIR3DL1 mutagenesis. We further document the clinical relevance of non-expressed vs. expressed KIR3DL1 alleles to relapse incidence in a local cohort of hHSCT patients.

Cohort of T-Replete Haploidentical HSCT Patients
This study analyzed a cohort of 186 adult patients with hematological malignancies who underwent T cell-replete haploidentical hematopoietic stem cell transplantation (hHSCT) with post-transplant cyclophosphamide (PTCy) in the Hematology Department of Nantes University Hospital. Various conditioning regimens were used, including a reduced-intensity TBF regimen [39], a Baltimore-based regimen [40][41][42], and a myeloablative or sequential regimen [40,43]. The source of grafts in all cases was peripheral blood stem cells from a haploidentical donor. Graft versus host prophylaxis consisted of PTCy, cyclosporine A, and mycophenolate mofetil for all cases. High-resolution typing for HLA-A, -B, and -C loci was carried out for all donor and recipient pairs by next-generation sequencing using Omixon Holotype HLA (Omixon, Budapest, Hungary). All patients and donors provided written informed consent for their data to be collected in the PROMISE database of the European Society for Blood and Marrow Transplantation. This study complied with the Declaration of Helsinki and was approved by the Ethics Review Board of Nantes University Hospital.
The clinical outcome and immune reconstitution of some patients have been previously reported [44,45] and were updated in December 2022 for this study. The main objective was to assess the impact of non-expressed vs. expressed donor KIR3DL1 allotype on relapse incidence.

KIR Genotyping
Generic KIR typing was performed for all blood donors (N = 278) and HSC donors (N = 186) using a KIR multiplex PCR-SSP method [46]. KIR genotypes and tel motifs were assigned as reported [8,47]. TelAA, telAB, and telBB KIR motifs in all blood donors were defined, taking into account KIR3DL1/S1/2DS1/2DS4 genes [8]. In particular, telAA individuals were characterized by the presence of KIR3DL1 and 2DS4 and the absence of KIR3DS1 and 2DS1 genes. TelAB individuals were characterized by the presence of KIR3DL1 and 2DS4 with 3DS1 and/or 2DS1 genes. TelBB individuals were characterized by the presence of KIR3DS1 and/or 2DS1 and the absence of KIR3DL1 and/or 2DS4 genes.

KIR Allele Typing
To assign KIR3DL1/S1 alleles in blood donors (N = 278) and HSC donors (N = 186), KIR genes were captured by long-range PCR and subjected to sequencing on a MiSeq sequencer (Illumina, San Diego, CA, USA) after library preparation as reported [38]. KIR3DL1/S1 allele assignment was performed by using Profiler software version 2.24, developed by M. Alizadeh (Research Laboratory, Blood Bank, Rennes, France) [38]. An updated KIR allele library, available on the IPD-KIR database, was implemented in Profiler. KIR3DL1/S1 allele combinations and corresponding tel motifs in blood donors are shown in Table S1.

KIR3DL1 Constructs
KIR3DL1 constructs were made from pcDEF3-3DL1*004 and pcDEF3-3DL1*002 (control) vectors, kindly provided by P. Parham (Stanford, CA, USA), in which enhanced green fluorescent protein (eGFP) was attached to the C terminus of KIR3DL1 (KIR3DL1-eGFP). Due to unexpected point mutations in the pcDEF3 vector and eGFP, a recombinant PCR approach was used to make chimeric KIR3DL1-eGFP constructs from the pcDEF3-3DL1*004 vector and targeted 3DL1 mutations described in the "Sitedirected mutagenesis of KIR3DL1" section. To generate the KIR3DL1-eGFP constructs, the NEBuilder Hifi DNA Assembly cloning kit with Q5 High-Fidelity DNA polymerase (New England Biolabs, Ipswich, MA, USA) was used. Amplification of KIR3DL1 was performed from the pcDEF3-KIR3DL1*004 vector with the sense KIR3DL1 primer (5cagatatccatcacactggcccaccatgtcgctcatggtcgtc-3 ), which overlaps the 3 end of pcDEF3, and the antisense KIR3DL1 primer (5 -tgctcaccattgggcaggagacaactttg-3 ), which overlaps the 5 end of eGFP. The amplification of eGFP was performed using the sense eGFP primer (5 -ctcctgcccaatggtgagcaagggcgag-3 ), which overlaps the 3 end of KIR3DL1, and the antisense eGFP primer (5 -acactatagaatagggccctttacttgtacagctcgtccatg-3 ). The template for eGFP amplification was the pcDEF3-KIR3DL1*004 vector. Overall, recombinant amplification was performed using the KIR3DL1 and eGFP amplicons as a template with forward KIR3DL1 and reverse eGFP primers and cloned into the pcDEF3 vector. The strategy used to generate KIR3DL1-eGFP constructs is shown in Figure S1. KIR3DL1-eGFP constructs were sequenced on an ABI 3730XL instrument (Eurofins Genomics, Ebersberg, Germany). Error-free KIR3DL1-eGFP clones were subcloned into the pcDEF3 expression vector.

Site-Directed Mutagenesis of KIR3DL1
Point mutations in the KIR3DL1*004-eGFP construct were generated using the GeneArt Site-Directed Mutagenesis System (Life Technologies) and oligonucleotide primers containing the relevant mutations, as recommended by the manufacturer. Position 152 of KIR3DL1*004 was changed from A to G, resulting in a Y30C amino acid change (KIR3DL1*019). Position 320 of KIR3DL1*019 was changed from C to T, resulting in an L86S amino acid substitution (KIR3DL1*019 L86S ). Position 607 of KIR3DL1*019 and KIR3DL1*019 L86S was changed from T to C, resulting in an S182P amino acid substitution (KIR3DL1*019 S182P and KIR3DL1*019 L86S+S182P , respectively). The full coding sequences of the resulting KIR3DL1*019, L1*019 L86S , L1*019 S182P , and L1*019 L86S+S182P plasmids were sequenced to confirm the mutations. KIR3DL1 constructs with their corresponding mutations are shown in Table 1.

Statistical Analyses
All statistical analyses were performed using R version 4.2.2 and GraphPad Prism v6.0 software (San Diego, CA, USA). Median follow-up was estimated with the reverse Kaplan-Meier method. Patient characteristics were compared using the chi-squared test for discrete variables and Student t-test for continuous variables. The clinical outcomes studied were overall survival (OS), defined as the probability of survival, and disease-free survival (DFS), defined as survival with no evidence of relapse, from day 0 of hHSCT. OS and DFS were compared using the log-rank test and Kaplain-Meier graphical representation. Relapse was calculated using cumulative incidence, considering non-relapse mortality (NRM) as a competing risk. Univariate and multivariate analyses were performed using the Cox proportional-hazard model. Factors with a p-value of <0.1 by univariate analysis or of interest for the study were included in multivariate analysis. A p-value of <0.05 was considered statistically significant.

Quality Management System (QMS)
All procedures were conducted under ISO9001:2015 and in compliance with the Guidance Document on Good In Vitro Method Practices.

Intracellular Localization of KIR3DL1*019
The KIR3DL1 + NK cell phenotype in L1*019-positive individuals suggests a poor expression of the putative protein on the NK cell surface, as reported for L1*004 [27]. Indeed, the expected mature L1*019 protein contains the same L86 and S182 amino acids involved in the intracellular retention of the L1*004 allotype [27], and only one amino acid in the D0 domain differs between L1*004 and L1*019 (Table 1).

Intracellular Localization of KIR3DL1*019
The KIR3DL1 + NK cell phenotype in L1*019-positive individuals suggests a poor expression of the putative protein on the NK cell surface, as reported for L1*004 [27]. Indeed, the expected mature L1*019 protein contains the same L86 and S182 amino acids involved in the intracellular retention of the L1*004 allotype [27], and only one amino acid in the D0 domain differs between L1*004 and L1*019 ( Table 1).
The absence of KIR3DL1 on the NK cell surface using DX9 or Z27 binding for L1*019positive individuals could be due to either a lack of recognition of epitopes targeted by these anti-KIR3DL1 mAbs or the translation of L1*019, impairing its expression on NK cells. To address these hypotheses, we prepared constructs encoding chimeric proteins in which enhanced GFP was attached to the C terminus of KIR3DL1 containing the coding sequence of L1*004 (no Z27 binding) or L1*002 (high Z27 binding). We performed site mutagenesis on L1*004 constructs to evaluate their impact on L1*019 expression ( Table 1). The Jurkat cell line was stably transfected with different KIR3DL1-eGFP constructs. Surface and intracellular KIR3DL1 expression were examined using flow cytometry and fluorescent microscopy focusing on eGFP+ cells, since the detection of eGFP was linked to the complete translation of the associated KIR3DL1. As expected, the Jurkat cell line transfected with L1*002 showed high Z27 surface staining and with L1*004 showed no surface staining (Figure 2a,b). The Jurkat cell line transfected with L1*019 also showed no Z27 surface binding (Figure 2a,b). Amino acid substitution at position 182 in the D1 domain ( S182P ) had no effect on Z27 surface binding (Figure 2a,b). In contrast, amino acid substitution at position 86 in the D0 domain alone ( L86S ) restored Z27 surface binding, and more significantly when coupled with amino acid substitution at position 182 in the D1 domain ( L86S+S182P ) (Figure 2a,b). The Jurkat cell line transfected with L1*002 or L1*004 showed intracellular expression of KIR3DL1, although lower expression was observed for L1*004, suggesting either lower binding with the 177,407 mAb or reduced intracellular expression linked to the specificity of this L1*004 allele (Figure 2c). The Jurkat cell line transfected with L1*019 showed intracellular binding comparable to L1*004 (Figure 2c). Amino acid substitution at position 182 in the D1 domain ( S182P ) partially increased the intracellular binding with the 177,407 anti-KIR3DL1 mAb (Figure 2c). Strikingly, amino acid substitution at position 86 in the D0 domain alone ( L86S ) or coupled with amino acid substitution at position 182 in the D1 domain ( L86S+S182P ) strongly increased intracellular 177,407 binding ( Figure 2c). Merged images showed co-localization of KIR3DL1 and HLA class I for L1*002, but not for L1*004 (Figure 2d), confirming the intracellular retention of L1*004 [27]. Intracellular staining using the 177,407 mAb revealed retention of L1*019 in the cytoplasm (Figure 2d). Concordant with the flow cytometry data, amino acid substitution at position 182 in the D1 domain ( S182P ) had no effect on KIR3DL1 intracellular expression (Figure 2d). the complete translation of the associated KIR3DL1. As expected, the Jurkat cell line transfected with L1*002 showed high Z27 surface staining and with L1*004 showed no surface staining (Figure 2a,b). The Jurkat cell line transfected with L1*019 also showed no Z27 surface binding (Figure 2a,b). Amino acid substitution at position 182 in the D1 domain ( S182P ) had no effect on Z27 surface binding (Figure 2a,b). In contrast, amino acid substitution at position 86 in the D0 domain alone ( L86S ) restored Z27 surface binding, and more significantly when coupled with amino acid substitution at position 182 in the D1 domain ( L86S+S182P ) (Figure 2a,b). The Jurkat cell line transfected with L1*002 or L1*004 showed intracellular expression of KIR3DL1, although lower expression was observed for L1*004, suggesting either lower binding with the 177,407 mAb or reduced intracellular expression linked to the specificity of this L1*004 allele (Figure 2c). The Jurkat cell line transfected with L1*019 showed intracellular binding comparable to L1*004 (Figure 2c). Amino acid substitution at position 182 in the D1 domain ( S182P ) partially increased the intracellular binding with the 177,407 anti-KIR3DL1 mAb (Figure 2c). Strikingly, amino acid substitution at position 86 in the D0 domain alone ( L86S ) or coupled with amino acid substitution at position 182 in the D1 domain ( L86S+S182P ) strongly increased intracellular 177,407 binding (Figure 2c). Merged images showed co-localization of KIR3DL1 and HLA class I for L1*002, but not for L1*004 (Figure 2d), confirming the intracellular retention of L1*004 [27]. Intracellular staining using the 177,407 mAb revealed retention of L1*019 in the cytoplasm (Figure 2d). Concordant with the flow cytometry data, amino acid substitution at position 182 in the D1 domain ( S182P ) had no effect on KIR3DL1 intracellular expression (Figure 2d).

Non-Expressed KIR3DL1 Alleles Are a Risk Factor for Relapse Incidence after T-Replete Haploidentical HSCT in Myeloid Diseases
Our functional data, obtained in vitro using dedicated KIR3DL1 constructs and NKL transfectants, suggest a putative role of KIR3DL1 null allotypes. We hypothesized that, in Only VMAPRTVLL leader peptide led membrane HLA-E expression and CD94/NKG2A binding. Statistical differences between groups were analyzed using one-way ANOVA; * p < 0.05.

Non-Expressed KIR3DL1 Alleles Are a Risk Factor for Relapse Incidence after T-Replete Haploidentical HSCT in Myeloid Diseases
Our functional data, obtained in vitro using dedicated KIR3DL1 constructs and NKL transfectants, suggest a putative role of KIR3DL1 null allotypes. We hypothesized that, in vivo, an inflammatory context, such as the cytokine storm observed early after hHSCT, could favor a specific environment suitable for stabilizing KIR3DL1 null allotypes on the NK cell surface. Moreover, we previously reported that KIR genetics impact hHSCT outcomes [44,45], but so far KIR3DL1 allele polymorphism has not been investigated in this context. Therefore, we investigated the impact of KIR3DL1 alleles in donors on the outcome of hHSCT (N = 186). Using our NGS technology, KIR3DL1 alleles were fully assigned in 166 HSC donors, including 22 donors with non-expressed KIR3DL1 (i.e., homozygous L1*004 or L1*004/3DS1, or L1 negative) and 144 donors with expressed KIR3DL1 (Figure 4a). The clinical characteristics of recipients were comparable in the two groups, based on donor KIR3DL1 expression (Table S2). The cumulative incidence of relapse after hHSCT in non-expressed vs. expressed KIR3DL1 allotypes did not differ when all patients were included (Figures 4b and S2a) or restricted to patients treated for a lymphoid neoplasm (Figures 4b and S2b). Interestingly, in patients treated for a myeloid malignancy, the cumulative incidence of relapse was significantly higher in those who received a graft with a non-expressed vs. expressed KIR3DL1 allotype (2 y relapse rate: 47 ± 14% vs. 24 ± 4%, p = 0.023) (Figure 4b,c). The same trend was observed in AML patients, although it was not significant due to a limited sample size (Figures 4b and S2c). A significant impact of donor KIR3DL1 allotypes on OS or DFS was not found. Based on the hypothesis that KIR3DL1 null can be expressed on NK cell surface following stress or stimulus, we evaluated KIR3DL1 null expression post-hHSCT (Figure 4d). KIR3DL1 expression on donor NK cells was recovered on recipient NK cells at days 30 and 60. KIR3DL1 is co-expressed with KIR2DL1/2/3. KIR3DL1 null is often co-expressed with KIR3DS1, the expression of which is increased after hHSCT, and co-expressed with KIR2DL1/2/3 (Figure 4d). Although NK cells are activated early after hHSCT, KIR3DL1 null is not expressed on NK cells, as observed in one representative KIR3DL1 null KIR3DS1 + HSC donor (Figure 4e).
To determine whether HSC grafts with non-expressed KIR3DL1 increase relapse incidence after hHSCT in patients with myeloid malignancies, univariate and multivariate analyses were performed considering confounding factors, such as age, diseases, DRI, status at the time of transplantation (first complete response (CR) versus subsequent CR versus lack of CR), and conditioning. In myeloid patients, univariate analysis identified recipient age (continuous), diseases (AML versus other myeloid diseases), DRI, status, conditioning, and donor non-expressed KIR3DL1 allotype as significant factors predicting relapse (Table 2). Multivariate analysis confirmed that age, diseases, status, and donor non-expressed KIR3DL1 allotype were associated with relapse after hHSCT in patients with myeloid disease (Table 2). Overall, these results sustain a deleterious effect of donor non-expressed KIR3DL1 alleles on relapse incidence after hHSCT only in the presence of myeloid malignancies. KIR3DL1 null expression post-hHSCT (Figure 4d). KIR3DL1 expression on donor NK cells was recovered on recipient NK cells at days 30 and 60. KIR3DL1 is co-expressed with KIR2DL1/2/3. KIR3DL1 null is often co-expressed with KIR3DS1, the expression of which is increased after hHSCT, and co-expressed with KIR2DL1/2/3 (Figure 4d). Although NK cells are activated early after hHSCT, KIR3DL1 null is not expressed on NK cells, as observed in one representative KIR3DL1 null KIR3DS1 + HSC donor (Figure 4e).  (d) Representative density plots of NK cells expressing KIR3DL1 and KIR3DS1 (Z27 mAb) coexpressed with KIR2DL1/2/3 (GL183 and EB6 mAbs) in graft and at days 30 and 60 post-hHSCT from one KIR3DL1 null KIR3DS1 + and one KIR3DL1 + KIR3DS1 − HSC donor. (e) Representative density plots of NK cells stained with KIR3DL1/3DS1 − and KIR3DL1-specific mAbs (Z27 and DX9 mAbs, respectively), leading to discrimination of KIR3DL1 and KIR3DS1 expression in graft and at days 30 and 60 post-hHSCT from one KIR3DL1 null KIR3DS1 + HSC donor. Frequency is indicated for each gate. KIR3DL1 alleles are classified depending on corresponding expression on NK cell surface (null, low, high) using a specific red (null) and blue/purple (expressed) color gradient code. * p < 0.05.

Discussion
The KIR3DL1/S1 allele frequencies established here are concordant with previous studies mainly performed in non-European populations [37,49,50]. In particular, we confirm the predominance of the non-expressed L1*004 allele and the scarcity of the KIR3DL1*019 allele in European populations [51,52]. Interestingly, the observed frequency of L1*004 is particularly high, reaching more than 20%. This is in agreement with Alicata et al. [14], who reported that around 20% of individuals had no expressed KIR3DL1 in a huge cohort of unrelated HSC donors from Italy. In addition to L1*004, we report frequencies ranging from 10 to 20% for L1*005, L1*001, L1*002, and S1*013. Of note, these KIR3DL1/S1 alleles are common in Europe [37]. In contrast, in non-European populations, as was recently described for Iranians [53], L1*001 and S1*013 allele frequencies reach near 50 and 30%, respectively, counterbalanced by a low frequency of L1*004 allele, around 5%. Overall, our mechanistic findings, established from French healthy blood donors, highlight a great diversity of KIR3DL1 allele polymorphism depending on telomeric KIR motifs. One could expect that this KIR3DL1 allele polymorphism could impact both NK cell phenotype and functions, as we previously reported for KIR2DL [45].
We showed that L1*019 is retained within the cell but is able to recognize some Bw4 ligands. Substitution of leucine for serine at position 86 in the D0 domain seems to be responsible for the poor folding of L1*019, with a minor contribution from position 182 in D1, as reported for L1*004 [27]. However, both substitutions (L86S and S182P) restored strong L1*019 expression on the membrane of the Jurkat cell line. Intracellular staining revealed poor L1*019 expression in the Jurkat cell line in comparison to the L1*004 allotype. Substitution (L86S) in D0 restored strong intracellular L1*019 staining similar to the L1*002 allotype. Position 30 in the D0 domain constitutes a unique divergent residue between L1*004 and L1*019 that may explain the difference in intracellular staining. We suggest that a functional interaction of KIR3DL1*019 L86S+S182P with some HLA-B allotypes expressed on transfected 221 target cells confers an inhibition of NKL degranulation. This result underlines the biological relevance of functional KIR3DL1 allotypes with intracellular localization. KIR3DL1*004 was previously shown to be involved in NK cell licensing, since the small amount reaching the surface can deliver inhibitory signals [28], although inefficient folding causes most of the L1*004 protein to be retained within the cell. The role of L1*019 in NK cell education remains unknown and needs further investigation. Altogether, these observations suggest a potential role for these KIR3DL1 allotypes in a pathological context. In the hHSCT context, NK cells are particularly activated; however, our investigation demonstrates an absence of KIR3DL1 null allotype expression on NK cells. We can hypothesize that self-induced or non-self proteins may interact and stabilize KIR3DL1 null allotypes on the cell surface, contributing to the NK cell response.
The relevance of polymorphic KIR genes remains a key component of NK cell-based immunotherapies for leukemic patients [54]. The rise of hHSCT in recent years also offers a privileged context to set a beneficial NK cell alloreactivity. KIR3DL1 has been previously correlated with HSCT outcomes [14,15,33], and we demonstrate for the first time that the KIR3DL1 polymorphism may play a pivotal role in relapse incidence after hHSCT. The deleterious effect of donor non-expressed KIR3DL1 allotypes on relapse incidence was only relevant in a limited cohort of patients with myeloid diseases confirming the lineage-specific relapse prediction after hHSCT [55]. Conversely, KIR3DL1*004 is reported to be protective against relapse for patients with AML after HLA-matched HSCT in a large amount of registry data [15]. The divergent effect of this common KIR3DL1 null allele between hHSCT-and HLA-matched HSCT could be related, in particular, to the HLA class I environment [14,56]. In our cohort, HLA-Bw4 environment of donors and recipients defined from both HLA-A and HLA-B typing has been assigned. We did not observe a significant impact of Bw4 environment and/or KIR3DL1/HLA-B subtype combinations on relapse incidence after hHSCT probably due to a limited size cohort. This should be investigated on a larger cohort. The divergent effect of the KIR3DL1*004 allele on HSCT outcome described by Boudreau et al. [15] and that reported here could also be due to the sample size and the heterogeneity concerning the proportion of AML patients. Indeed, the deleterious effect of non-expressed KIR3DL1 alleles on relapse incidence we reported after hHSCT, was observed in a limited cohort of patients with various myeloid diseases. Nonetheless, further investigations on a broader cohort of hHSCT restricted to AML patients would be necessary. In contrast to previous studies [55,57,58], the reported beneficial effect of inhibitory KIR ligand mismatches on relapse incidence after hHSCT did not reach significance here. Moreover, the protective effect of donor cenAA KIR motifs on relapse incidence that we observed in a limited cohort of hHSCT patients [45] was not confirmed, although a trend of less relapse in cenAA than cenB+ donors was observed in patients with myeloid diseases. Differences in the proportions of myeloid patients, conditioning regimens, and stem cell sources between published studies and what is reported here could explain these discordances. In addition to KIR3DL1, the lack of CR was the most significant factor affecting relapse incidence post-hHSCT, as expected [45,59,60].
For patients lacking Bw4, KIR3DL1-expressing NK cells from Bw4+ donors could be alloreactive following hHSCT. Given the predominance of the KIR3DL1*004 allele, the KIR3DL1 + NK cell repertoire post-hHSCT could be skewed. The lack of KIR3DL1 expression on NK cells could be associated with an over-representation of the KIR2DL + NK cell compartment, a possibility that needs further investigation. Indeed, donor KIR3DL1 + and KIR2DL + NK cell recovery at day 30 post-hHSCT was inversely impacted by KIR ligand mismatches [44]. More broadly, other KIR allele polymorphisms besides KIR3DL1 allotypes that also impact NK cell phenotype and function, such as KIR2DL and KIR2DS4, may be involved after hHSCT and should be investigated in a larger cohort.

Conclusions
Deciphering KIR allele polymorphism to better characterize the structure of the functional NK cell repertoire remains a significant challenge. The influence of KIR alleles on hHSCT outcomes is still poorly understood. We might expect that knowledge of how KIR allele distributions depend on KIR gene content could help in defining an algorithm to better select haploidentical donors, as reported in HLA-matched unrelated HSCT [8], as well as promoting a beneficial anti-leukemic effect driven by NK cells. More broadly, including KIR allele polymorphism could pave the way to improving our understanding of heterogeneous NK cell responses against acute leukemia [61] and the efficiency of NK-cell-based immunotherapies.