Impact of Graft-Versus-Graft Natural Killer Cell Alloreactivity on Single Unit Dominance After Double Umbilical Cord Blood Transplantation

Abstract

Background 

Natural killer (NK) cell alloreactivity is favored after double umbilical cord blood transplantation (dUCBT) in which cord blood (UCB) units and patients are often HLA class I mismatched. Generally, only 1 UCB unit persists after dUCBT. We hypothesize, that NK cell alloreactivity mediated by killer cell immunoglobulin-like receptor (KIR)-HLA interactions may explain the dominance of 1UCB unit over the other after dUCBT.

Methods 

We investigated the impact of KIR⁺ NK cell alloreactivities on the dominance of 1 full UCB unit in 50 dUCBT. We analyzed the effects of the KIR/HLA genetic incompatibilities and studied cord blood cells at both the phenotypic and functional levels.

Results 

The genetic combination of KIR3DL1⁺ loser UCB unit/Bw4⁻ winner UCB unit determined both the dominance of 1 UCB unit (hazards ratio, 2.88 [1.32-6.27], P = 0.0077) and correlated with an increased incidence of relapse (hazards ratio, 4.91 [1.39-17.3], P = 0.0134). It is interesting to note that cord blood cells exhibited extremely low HLA class I expression. Moreover, resting cord blood KIR3DL1⁺ NK cells exhibited a basal alloreactivity against Bw4⁻ target cells that increased upon activation, thus triggering death by apoptosis.

Conclusions 

Our unicentric study suggests, for the first time, the significant impact of KIR⁺ NK cell alloreactivity in the determination of which UCB unit will dominate in dUCBT.

In Brief

In this study, the authors try to understand why, in case of double umbilical cord blood transplantation, only one cord blood unit persists. In 50 patients that received a double unit cord blood transplantation, they suggest that KIR+ NK cell alloreactivity is mainly responsible for the loss of one unit.

Umbilical cord blood transplantation (UCBT) is an acceptable alternative for hematopoietic stem cell transplantation (HSCT), when fully HLA-matched donors are unavailable.^1,2 Compared with other hematopoietic stem cell (HSC) sources, UCBT is associated with a lower incidence of acute graft-versus-host disease (aGvHD) and relapse. However, the outcome of UCBT often results in delayed engraftment and a lower overall probability of engraftment.³ The choice of umbilical cord blood (UCB) unit(s) depends on the number of total nucleated cells (TNC), CD34⁺ stem cell count, and on UCB unit/patient HLA matching.^4-10 Adult recipients require 2 UCB units (double UCBT [dUCBT]) in order to achieve sufficient numbers of stem cells for infusion.^11-13 Interestingly, when full chimerism is obtained after dUCBT, it is usually derived from only 1 of the 2 UCB units infused. Neither the total number of TNC or CD34⁺ stem cells in the UCB unit, nor the level of HLA-A, -B, and -DRB1 matching enable systematic prediction of which 1 of the 2 UCB units will engraft.^5,14 In this “ménage à trois,” both T and/or natural killer (NK) cell alloreactivities between UCB units and patient in the graft-versus-host (GvH) or the host-versus-graft (HvG) directions, and between the 2 UCB units in the graft-versus-graft (GvG) direction should be considered.¹⁵ In this regard, single UCB unit dominance was shown to correlate with a functional CD8⁺ T cell response against the nonengrafted UCB unit.¹⁶ In contrast to T lymphocytes, there is a rapid recovery of NK cells after UCBT.^17,18 These cells represent the only lymphoid cell population potentially able to control leukemic relapse in the months preceding T cell reconstitution. Indeed, in extensively T cell–depleted HSCT, as is the case for haploidentical grafts with HLA class I–mismatched donor/recipient pairs, alloreactive NK cells may exert a beneficial graft-versus-leukemia effect.¹⁹ A beneficial effect of alloreactive NK cells in T cell–repleted haploidentical HSCT, recently reintroduced with postgraft infusion of high-dose cyclophosphamide, has been reported.²⁰ NK cells are tightly regulated by a complex repertoire of cell surface receptors controlling proliferation and activation that could be evaluated by phosphorylated P38 intracellular expression²¹ and potentially leading to apoptosis, cytokine production, and cytotolytic potential evaluated by CD107a mobilization assay. The effector functions of NK cells are regulated in particular by inhibitory and activating receptors; that is, killer cell immunoglobulin-like receptors (KIR), which recognize specific HLA-class I molecules (referred to as KIR ligands).²² In particular, HLA-Cw allotypes with Asn80 (C1 ligands) or with Lys80 (C2 ligands) are recognized by KIR2DL2/2DL3 and KIR2DL1, respectively. Note that KIR2DL2/3 recognizes a larger spectrum of HLA-C antigens than KIR2DL1.^23-25 HLA-A and -B allotypes, with a Bw4 motif, are targeted by KIR3DL1, whereas HLA-A3/A11 antigens are recognized by KIR3DL2.^26,27 Lack of inhibitory KIR engagement can trigger alloreactive KIR⁺ NK cell cytotoxicity only within functionally competent NK cells, that is, cells which have undergone “licensing” upon prior engagement of their inhibitory KIRs with self-HLA class I molecules during maturation.²⁸ Although the ligands and functions of inhibitory KIRs (KIR2DL, KIR3DL) are well documented, this is not the case for activating KIR (KIR2DS, KIR3DS1), except for KIR2DS1 and KIR2DS2 which, respectively, recognize C2 and C1 ligands as their inhibitory counterparts.^25,29

T cell and/or NK cell alloreactivity are particularly favored in a dUCBT context in which both UCB units and the patient may be mismatched for HLA-A, -B, -C, or -DRB1 loci that might in turn impact dUCBT outcome. We recently demonstrated that KIR⁺ cord blood NK cells reconstitute the patient's hematopoiesis as soon as day +14 post-dUCBT.³⁰ However, there have been reports that in UCBT KIR ligand disparities (in the GvH direction) are beneficial,³¹ deleterious,³² or are without any significant effect.³³ These discordant results may be dependant on the chosen clinical endpoint, or on which graft parameters are studied. Moreover, in these studies, neither KIR genes, especially activating ones, nor KIR/KIR ligand combinations have been analyzed for their impact on NK alloreactivity and transplant outcome. Sekine et al³⁴ reported a beneficial effect of some KIR2D/HLA-C genetic combinations in the GvH direction on UCBT outcome. The impact of KIR/KIR ligand combinations in the GvG direction was only evaluated in 1 unicentric study,³⁵ which reported no identifiable role of KIR/HLA genotypes on UCB dominance after dUCBT. Recently, we confirmed, from a large dUCBT cohort, that included patients from several French transplant centers, that KIR/KIR ligand genetic combinations have no significant impact on UCB dominance.³⁶ However, extended KIR/KIR ligand genetic analysis using more homogeneous dUCBT cohorts is needed to determine prospective use of KIR/HLA combinations in the selection of UCBs that will improve engraftment and immune reconstitution. Therefore, we investigated the contribution of both HLA and KIR genetic disparities on 1 full UCB unit dominance in an unicentric dUCBT cohort. To better define the impact of our genetic results, and to propose a mechanistic explanation, we investigated cord blood cells at both the phenotypic and functional levels.

MATERIALS AND METHODS

Patient, Disease, and Transplant Characteristics

This retrospective study first included 68 consecutive patients who received dUCBT at Nantes CHU between March 2006 and December 2011. Ten patients without available DNA material for both UCB units or incomplete hematopoietic chimerism data and 8 patients with primary engraftment failure were excluded. Overall, 50 patients who received dUCBT and engrafted with 1 full dominant UCB unit were recruited at a single hematological department (Table S1, SDC,https://links.lww.com/TP/B370). All patients provided written informed consent in accordance with the Helsinki Declaration before UCBT. The present study obtained approval from the University of Nantes Institutional Review Board.

Clinical Endpoint

The main clinical end-point was engraftment with 1 full dominant UCB unit.³⁷ Engraftment was defined as the incidence of neutrophil recovery (≥0.5 × 10⁹/L) in association with full chimerism by 42 days after transplantation. Full chimerism was defined as greater than 90% of donor cells.^14,38 The predominant UCB unit and the nonpredominant UCB unit in engrafted patients were called the “winner” and the “loser” UCB unit, respectively. aGvHD was evaluated according to the modified Glucksberg grading system.³⁹

Evaluation of Hematopoietic Chimerism

Hematopoietic chimerism data were measured for the first 3 months after dUCBT and were evaluated using real-time quantitative polymerase chain reaction of genetic markers specific for the patient and for the UCB units as previously described.^30,40

Cells (Peripheral Blood Mononuclear Cells, Cord Blood Samples, and Cell Lines)

For phenotypic and/or functional assays, peripheral blood mononuclear cells (PBMC) were isolated from healthy adult volunteers. Cord blood cells were isolated from umbilical cord blood samples. All blood donors were recruited at the Blood Transfusion Center (EFS, Nantes, France) and umbilical cord blood samples were obtained at the Nantes CHU maternity unit. Informed consent was obtained from all healthy individuals and mothers. HLA class-I deficient 721.221 lymphoblastoid Epstein-Barr virus-B cells, referred to as 221 cells, Bw4 (B*15:13) - and Bw6 (B*39:01)-transfected 221 cells (respectively named 221-Bw4 and 221-Bw6) were used to assess NK cell degranulation. Two different Epstein-Barr virus-B cell lines were used as target cells to amplify NK cells as we previously described.²⁹

HLA Genotyping and UCB Unit Selection

Intermediate (HLA-A, -B) or high (HLA-DRB1) resolution typing was performed by reverse Sequence-Specific-Oligoprobe (SSO) Luminex technology (One Lambda, Inc., Canoga Park, CA) for all patients and UCB units. High resolution typing for HLA-C was carried out retrospectively on all UCB unit/patient pairs by Sequence Based Typing kit (Abbott Molecular Park, IL). All UCB units were matched to the recipient for at least 4 out of the 6 HLA-A, -B, -DRB1 alleles based on antigen-level (HLA-A, -B), and allele level HLA-DRB1 typing and to each other at 4 (or more) of the 6 HLA antigens, though not necessarily at the same locus.

KIR Genotyping

All patients, UCB units, cord blood samples and healthy individuals were typed for KIR2DL1, 2DL2, 2DL3, 2DL4, 2DL5, 3DL1, 3DL2, 3DL3, 2DS1, 2DS2, 2DS3, 2DS4/1D, 2DS5, 3DS1 using either a KIR genotyping polymerase chain reaction-sequence specific primer kit (Invitrogen, Compiègne, France) or a multiplex polymerase chain reaction-sequence specific primer method as previously described.⁴¹

Potential T Cell and KIR⁺ NK Cell Alloreactivity

T cell alloreactivity was evaluated depending on HLA-A, -B, -C, and -DRB1 mismatches. KIR⁺ NK cell alloreactivity was evaluated depending either on the presence of a specific inhibitory KIR gene and the absence of its corresponding KIR ligand, or the presence of a specific activating KIR gene and the presence of its KIR ligand. In particular, KIR2DL1/2DL2/2DL3/3DL1⁺ NK cell alloreactivity was determined by the presence of KIR2DL1, KIR2DL2/L3 or KIR3DL1 genes with the absence of their KIR ligands for 2DL1 (C2), 2DL2/L3 (C1) and 3DL1 (Bw4) determined by HLA class I typing. KIR2DS1⁺ NK cell alloreactivity was dependant on the presence of the KIR2DS1 gene and the presence of its ligand (C2) (See Methods, SDC,https://links.lww.com/TP/B370).

RESULTS

Analysis of Graft Parameters Likely to Impact 1 Full UCB Unit Dominance After dUCBT

Fifty patients engrafted with 1 full dominant UCB unit after dUCBT were included in this study (Table S1, SDC,https://links.lww.com/TP/B370). Neutrophil recovery occurred at a median of 19 days [range 8–60]. TNC dose, CD34⁺ stem cell dose infused after thawing in each UCB unit, and the degree of HLA-A, HLA-B, HLA-DRB1 matching among the UCB units and patients had no significant impact on the dominance of 1 UCB unit after dUCBT (Table S1, SDC,https://links.lww.com/TP/B370).

No Impact of Potential T Cell Alloreactivity Based on HLA Mismatches on 1 Full UCB Unit Dominance After dUCBT

The impact of CD8⁺ T cell alloreactivity on UCB unit dominance after dUCBT has been previously reported.¹⁶ By the analysis of HLA mismatches among patient and UCB units, we looked at whether the potential alloreactivity of T lymphocytes would determine which UCB unit became dominant (Figure 1A). No significant correlation between 1 UCB unit dominance and the frequencies of HLA-A, -B, -C and -DRB1 mismatches in both GvH and GvG directions (Figure 1B) were observed in our series. Because multiple HLA class I mismatches were more frequent than single HLA class I mismatches (Figure S1, SDC,https://links.lww.com/TP/B370), the specificity of HLA class I molecules present in patients and UCB units was further documented.

FIGURE 1: HLA mismatches, KIR ligand and KIR gene frequencies among patients, loser and winner UCB units (A) Potential T cell and NK cell alloreactivity on 1 full UCB unit dominance after dUCBT in GvH, HvG, and GvG directions. B, HLA class I and HLA-DRB1 mismatches between winner UCB units (n = 50), loser UCB units (n = 50) and patients (n = 50) in both GvH and GvG directions. C, HLA-A3, A11+ (KIR3DL2), HLA-A and HLA-B Bw4+ (KIR3DL1), HLA-C1+ (KIR2DL2/L3/S2) and HLA-C2+ (KIR2DL1/S1) KIR ligand frequencies on winner UCB units (n = 50), loser UCB units (n = 50), patients (n = 50) and healthy blood donors (n = 87). D, Inhibitory (2DL/3DL) and activating (2DS/3DS) KIR gene frequencies established on winner UCB units (n = 50), loser UCB units (n = 50), patients (n = 50) and healthy blood donors (n = 87). Numbers and percentages were compared between groups with χ² test or Fisher exact test when appropriate.

KIR Ligand Distribution Between Both UCB Units and Patient

KIR NK cell alloreactivity is expected when NK cells in 1 UCB unit display a particular inhibitory KIR and its HLA class I ligand is absent in the patient (GvH direction) or in the second UCB unit (GvG direction) (Figure 1A). Potential KIR NK cell alloreactivity could also be expected when NK cells in 1 UCB unit have activating KIR gene(s) directed against known or putative HLA class I ligands expressed on patient cells, or by the second UCB unit (Figure 1A). We showed that the frequencies of different KIR ligands were similar between winner and loser UCB units as well as between patients and controls (Figure 1C).

Potential Alloreactivity of KIR3DL1⁺ NK Cells From the Loser UCB Toward the Bw4⁻ Winner UCB Unit

To evaluate a potential NK cell alloreactivity mediated by KIR, we focused our study on particular KIR/KIR ligand incompatibilities. Both inhibitory and activating KIR gene frequencies were similar in winner vs loser UCB units, as well as in patient vs controls (Figure 1D). Moreover, there was no significant correlation between 1 full UCB unit dominance and the inhibitory KIR/KIR ligand mismatches involving KIR2DL1/2/3/3DL2, or activating KIR/KIR ligand matches involving KIR2DS1/2/3/4/5/3DS1 genes and their respective or putative ligands (Figure 2A and data not shown). We found that the loser UCB unit expresses KIR3DL1, whereas the winner UCB unit does not have the corresponding Bw4 ligand. This finding was significantly higher in the GvG direction (26% vs 12%, P = 0.020, n = 13 vs n = 6, Figure 2B). Detailed analysis of the nature (Figure 2C) and the number (Figure 2D) of KIR/KIR ligand incompatibilities reveal similar patterns in the GvH direction, but some differences in the GvG direction also were noted.

FIGURE 2: Impact of KIR/KIR ligand genetic combinations on 1 full UCB unit dominance after dUCBT. A, The number of KIR2DL1⁺/C2⁻, KIR2DL2/L3⁺/C1⁻ and KIR2DS1⁺/C2⁺ genetic combinations between winner UCB unit/patient pairs (n = 50) and loser UCB unit/patient pairs (n = 50) in the GvH direction, and between winner UCB unit/loser UCB unit pairs (n = 50) and loser UCB unit/winner UCB unit pairs (n = 50) in the GvG direction was compared. B, The number of KIR3DL1⁺/Bw4⁻ genetic combinations between winner UCB unit/patient pairs (n = 50) and loser UCB unit/patient pairs (n = 50) in the GvH direction, and between winner UCB unit/loser UCB unit pairs (n = 50) and loser UCB unit/winner UCB unit pairs (n = 50) in the GvG direction was compared. C, Charts representing the detailed KIR/KIR ligand combinations taking into account KIR2DL1, KIR2DL2/L3, KIR3DL1 and KIR2DS1 genes and corresponding ligands in both the GvH and GvG directions. D, Charts representing the number of KIR. KIR ligand combinations in both the GvH and GvG directions taking into account isolated KIR2DL1⁺/C2⁻, KIR2DL2/L3⁺/C1⁻, KIR3DL1⁺/Bw4⁻ and KIR2DS1⁺/C2⁺ incompatibilities and cumulated ones. Numbers and percentages were compared between groups with χ² test or Fisher exact test when appropriate.

In both univariate and multivariate analyses, we confirmed that KIR3DL1⁺ loser UCB/Bw4⁻ winner UCB genetic combination was associated with 1 full UCB unit dominance (Table 1). This genetic combination remains significantly associated with neutrophil recovery in stratified populations (Table S2, SDC,https://links.lww.com/TP/B370), and in patients engrafted with 1 full dominant UCB unit and still living at 90 days post-dUCBT (Table S3, SDC,https://links.lww.com/TP/B370).

TABLE 1: Analysis of risk factors influencing neutrophil recovery after dUCBT (n = 50) with one full UCB unit dominance using Cox model

KIR3DL1⁺ Loser UCB/Bw4⁻ Winner UCB Genetic Combination Impacts on UCB Dominance and Relapse Incidence After dUCBT

We then evaluated the impact of this genetic combination on the time to neutrophil recovery, relapse, overall survival, and incidence of aGvHD. Interestingly, the time to neutrophil recovery of the KIR3DL1⁺ loser UCB/Bw4⁻ winner UCB pairs was significantly shortened compared with other dUCBT without this genetic combination (median 15[9-27] vs 24[6-60] days, P = 0.001, Figure 3A). The KIR3DL1⁺ loser UCB/Bw4⁻ winner UCB genetic combination was significantly associated with an increased incidence of relapse (P = 0.017, Figure 3B). The increase in the incidence of relapse was confirmed by multivariate Cox analysis (hazards ratio, 4.91 [1.39-17.3], P = 0.0134, data not shown). The KIR3DL1⁺ loser UCB/Bw4⁻ winner UCB genetic combination has no significant effect on overall survival (Figure 3C) and aGvHD incidence (Figure 3D).

FIGURE 3: KIR3DL1⁺ loser UCB/Bw4⁻ winner UCB genetic combination is associated with shortened neutrophil recovery but increased relapse incidence after dUCBT. Time to neutrophil recovery (A), cumulative incidence of relapse (B), cumulative incidence of overall survival (C), and cumulative incidence of aGvHD (D) in dUCBT with expressed KIR3DL1⁺ loser UCB/Bw4⁻ winner UCB unit genetic combination and without this genetic combination for all patients.

Low Expression of HLA Class I Molecules on Cord Blood Cells

Next, we investigated whether the lower expression of HLA class I antigens on cord blood cells, in comparison with adult PBMC, would have any effect on alloreactivity in the GvG direction, and not in the GvH direction. Indeed, HLA class I expression was significantly decreased on freshly isolated cord blood cells (MFI median = 682) compared with adult PBMC (MFI median, 1263; P = 0.0002, Figure 4A). Because dUCBT are performed using thawed UCB units, this difference was also confirmed on thawed cord blood cells (P = 0.0003, Figure 4A).

FIGURE 4: Umbilical cord blood cells are characterized by a low HLA class I expression and cord blood KIR3DL1⁺ NK cells are stimulated as a result of GvG interactions, resulting in their self-destruction (A) Scatter plots representing the MFI of HLA class I expression on freshly isolated (n = 15) and thawed (n = 8) cord blood cells, and freshly isolated (n = 11) and thawed (n = 18) adult PBMC. Results are expressed as the subtraction of mean intensity fluorescence of IgG control to mean intensity fluorescence of HLA class I molecules. Statistical significance between the 2 groups was determined using the 1-way ANOVA test. B, Representation of pP38 mean intensity fluorescence on cord blood (n = 4, dotted line) and adult (n = 6, full line) NK cells. Histograms illustrating the MFI of pP38 phosphoproteins on cord blood and adult NK cells (black line) compared to medium (grey filled line) are shown for 1 representing experiment. Statistical significance between 2 groups was determined using the 1-way ANOVA test. C, Summary box and whisker plot summarizing the percentages of CD107a⁺ KIR3DL1⁺ NKG2A⁻ NK cells in different stimulation conditions: medium, 221, 221-Bw4 cell lines for experiments performed from 7 Bw4⁺ cord blood samples. Top and bottom whiskers represent values of the top and bottom 25% of cases, respectively; boxed area, interquartile range. D, Summary box and whisker plot summarizing the percentages of CD107a⁺ NK and (E) CD107a⁺ KIR3DL1⁺ NK cells for experiments performed from in vitro amplified cord blood cells (n = 16) in different stimulation conditions: medium, 221 and 221-Bw4 cell lines at an NK/target ratio of 1:1. Statistical significance between 2 groups was determined using the 1-way ANOVA test. F, Scatter plots representing annexin-V expression of KIR3DL1⁺ cord blood NK cells against 221 and 221-Bw4 cell lines after in vitro amplification. Statistical significance between 2 groups was determined using the 1-way ANOVA test. G, Model representing the characteristics of both the winner and the loser UCB units based on KIR/HLA incompatibilities in the GvG direction. ANOVA, analysis of variance.

Cord Blood NK Cells Are Activated by P38 Pathway

We further evaluated the activation of cord blood NK cells in the context of dUCBT. Trotta et al²¹ highlighted the role of ERK and P38 MAPK kinases but not JNK MAPK in NK-cell mediated cytotoxicity. ERK pathway is similarly activated in cord blood and adult NK cells for the highest dose of PMA/ionomycin (data not shown). As illustrated in Figure 4B, P38 phosphorylation is high in cord blood NK cells (P = 0.032 and P = 0.023, Figure 4B), underlying the susceptibility of cord blood NK cells to be activated by the P38 pathway.

Cord Blood Units Exhibit a KIR3DL1⁺ NK Cell Alloreactivity Against Bw4⁻ Target Cells

To evaluate the activation of cord blood KIR3DL1⁺ NK cells, we first determined the potential alloreactivity of resting and activated cord blood NK cells with different target cells positive or negative for the expression of the Bw4 ligand. The degranulation CD107a profile of unmanipulated KIR3DL1⁺ NKG2A⁻ cord blood NK cells indicated a decreased degranulation of KIR3DL1⁺ NK cells when incubated with Bw4⁺ compared to 221 cell line (P = 0.002, Figure 4C). We further evaluated the degranulation of KIR3DL1⁺ NK cells from stimulated cord blood PBMC as previously described²⁹ to mimic the strong NK cell expansion observed early after dUCBT. We demonstrated there was alloreactivity of activated NK cells against the 221 cell line in a degranulation assay (P = 2.79 × 10⁻¹⁸, Figure 4D). Activated cord blood KIR3DL1⁺ NK cells showed strong alloreactivity against the 221 cell line compared to the 221-Bw4 cell line (P = 1.48 × 10⁻²¹ and P = 1.70 × 10⁻¹⁷, Figure 4E).

Death of Activated Cord Blood NK Cells by Apoptosis

Finally, we investigated if activation of cord blood NK cells can result in premature death by apoptosis. We therefore evaluated Annexin V expression of stimulated KIR3DL1⁺ cord blood NK cells against 221-Bw4⁺ or 221-Bw4⁻ targets. Interestingly, we observed a high Annexin-V expression of KIR3DL1⁺ NK cells against 221-targets as all NK cells (data not shown) and a protection against 221-Bw4⁺ targets (P < 0.0001) (Figure 4F).

DISCUSSION

In our study, we investigated the potential impact of both T cell and NK cell alloreactivities on 1 full UCB unit dominance after dUCBT by studying the extent of KIR/KIR ligand genetic incompatibilities between UCB units and patients. We correlated the phenotype to the function of cord blood KIR⁺ NK cells. In this unicentric cohort of dUCBT, excluding patients with mixed chimerism and engraftment failure, only the presence of the KIR3DL1 gene in the loser UCB unit and the absence of the corresponding Bw4 ligand in the winner UCB unit were significantly associated with an increased rate of neutrophil recovery, and with a higher incidence of relapse. Supporting our KIR/KIR ligand genetic data, we reported a lower expression of HLA class I molecules on cord blood cells compared with adult PBMC that could favor a cord blood KIR⁺ NK cell alloreactivity of 1 UCB unit against the second one rather than against the patient. Moreover, the absence of Bw4 ligand on winner UCB cells determines the direction of this NK cell alloreactivity between both UCB units.

Our genetic data are in agreement with previous studies that showed no significant impact of KIR ligand mismatches in the GvH direction on UCBT engraftment.^31,32,42 The data from Willemze et al,³¹ including patients undergoing sUCBT, showed that inhibitory KIR-ligand incompatibilities are associated with a decreased incidence of relapse, but had no significant impact on engraftment. In contrast, Brunstein et al³² reported a detrimental effect of KIR-ligand incompatibility on the increased incidence of relapse in recipients of sUCBT or dUCBT after a reduced intensity conditioning regimen but, the impact of these KIR ligand incompatibilities on engraftment was not presented. Lastly, Garfall et al⁴² reported no significant impact of KIR ligand incompatibility on dUCBT engraftment. However, our results extend the findings of these studies, as we highlight the impact of KIR/KIR ligand incompatibilities on engraftment, especially in the GvG direction. Tarek et al³⁵ reported no identifiable role of HLA and KIR genotypes in single unit dominance after dUCBT. However, their genetic analysis was focused on the distribution of AA/Bx KIR haplotypes, centromeric/telomeric motifs and the number of missing KIR ligands within UCB units and patient. Contributing to novel information for the field, our study evaluated individual inhibitory and activating KIR/KIR ligand combinations in both the GvH and GvG directions.

In our series, the majority of dUCBT evolve in a Bw4⁺ environment. Of note, cord blood KIR3DL1⁺/Bw4⁻ NK cells can acquire functional capacities firstly via Bw4⁺ cells from UCB units or patients and secondly via other inhibitory NK receptors⁴³ or massive cytokine production and cell contacts.⁴⁴ Indeed, NK cells with inhibitory KIR for nonself HLA class I can lyse target cells lacking cognate ligand after T cell–depleted HLA identical HSCT suggesting a rapid breaking of tolerance to self after either bone marrow or peripheral stem cell transplantation.⁴⁵ However, NK cells become progressively tolerized to self during the first 3 months post-HSCT.⁴⁶ Our results also suggest that activated cord blood KIR3DL1⁺ NK cells exhibit a strong alloreactivity in a dUCBT context leading to potential cell death by apoptosis. Overall, both our genetic and cellular data suggest that KIR⁺ NK cell–mediated immune interactions between both UCB units rather than T cell and/or intrinsic properties of each infused UCB unit may explain single-unit dominance observed after dUCBT. These results are also corroborated with data from Eldjerou et al⁴⁷ who showed that UCB unit dominance is an in vivo phenomenon associated with a GvG immune interaction mediated with CD34⁻ cells. No significant correlation between 1 UCB unit dominance and the frequencies of HLA-A, -B, -C, and -DRB1 mismatches in GvH and GvG directions was observed in our series. However, we cannot exclude the presence of functional alloreactive CD8⁺ T cells from winner UCB units against loser UCB units as described.¹⁶

The KIR3DL1⁺ loser UCB/Bw4⁻ winner UCB unit genetic combination was significantly associated with 1 full UCB unit dominance after dUCBT, although some other KIR2D/KIR ligand incompatibilities differed between winner and loser UCB units. The KIR3DL1⁺ loser/Bw4⁻ winner UCB unit combination was predominant notably due to the attenuated effects mediated by other KIR2DL receptors which share HLA-C ligands,⁴³ and by the frequency of HLA-Bw4⁺ in loser UCB units and patients, which were particularly high in our cohort compared with winner UCB units. Moreover, the ligands of KIR3DL1 were clearly identified in contrast to KIR2DL2/2DL3, which interacts with a larger spectrum of HLA-C molecules.²⁵ Overall, one could expect that other KIR/KIR ligand incompatibilities, such as KIR2D/HLA-C,³⁴ may impact dUCBT outcome depending on graft parameters, such as the level of HLA class I matching between both UCB units and patient, the disease and the ethnicity of the patient and UCB units. Of note, KIR gene/genotype and HLA class I allele frequencies differ between populations as observed for the distribution of C1, C2 and Bw4 ligands between Japanese and white populations.⁴⁸ Lastly, the large KIR allelic polymorphism such as reported for KIR3DL1 gene⁴⁹ may also affect both the phenotype and the function of cord blood KIR⁺ NK cells. Thus, KIR allelic polymorphism, contribution of other KIR genes and corresponding KIR ligands should be further investigated on a larger homogenous cohort in order to determine the hierarchy between all KIR/KIR ligand combinations functionally implicated in dUCBT outcome.

In addition to 1 full UCB unit dominance, the KIR3DL1⁺/Bw4⁻ winner UCB genetic combination was also significantly associated with a higher incidence of relapse in our series, but had no significant effect on overall survival and nonrelapse mortality (data not shown), suggesting that relapsed patients all get salvaged. This may also reflect a poor NK cell alloreactivity of the winner UCB unit which needs to be further investigated because functional NK cells able to kill acute myeloid leukemia blasts after UCBT have been reported.⁵⁰

Based on our results showing the significant impact of KIR3DL1⁺ loser UCB/Bw4⁻ winner UCB genetic combination on both the neutrophil recovery and relapse incidence, we suggest that strong alloreactivity of cord blood NK cells hampers their own engraftment and favors engraftment of the UCB unit which may exhibit a poor antileukemic NK cell alloreactivity. Taken together, we suggest that the winner UCB unit displays KIR/HLA features favoring its engraftment; for example, low KIR/HLA incompatibilities in the GvG direction, and low NK cell activation (Figure 4G). The loser UCB unit, which is characterized by high KIR/HLA incompatibilities and subsequent high NK cell activation may be responsible for apoptosis (Figure 4G). This GvG effect is also supported by the fact that NK cell licensing after dUCBT is driven by the self-HLA class I molecules from the winner UCB unit.⁵¹ Overall, our results obtained from a local dUCBT cohort suggest that it may prove advantageous to include KIR gene/KIR ligand content in the selection of UCB units, in addition to conventional tools such as UCB cell doses and HLA matching, to find the best UCB unit pairs, which can improve both engraftment, and decrease incidence of relapse after dUCBT.⁵²

ACKNOWLEDGMENTS

The authors would like to thank Patricia Herry, Dr. Anne Devys (Laboratoire HLA, EFS Nantes), Florent Malard (CHU Hotel Dieu, Nantes), Zakia Djaoud (EA4271, EFS Nantes), Pr. Véronique Sébille (EA4275 Université de Nantes), Pr. Mohamad Mohty (Hopital Saint Antoine, AP-HP Paris) for their collaboration, Pr. Matthew Albert (Inserm U818, Institut Pasteur, Paris) for having critically reviewed the manuscript and Britt House Europe (Hegenheim, France) for help in the editing of the article.